From owner-reliable_computing [at] interval [dot] usl.edu Tue Apr 3 11:35:31 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id LAA10222 for reliable_computing-outgoing; Tue, 3 Apr 2001 11:35:31 -0500 (CDT) Received: from Morgana.Elet.PoliMi.IT (root [at] morgana [dot] elet.polimi.it [131.175.21.1]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id LAA10217 for ; Tue, 3 Apr 2001 11:35:21 -0500 (CDT) Received: from host calvin and sender pc-pez.elet.polimi.it [131.175.70.208]; by mail relay Morgana.Elet.PoliMi.IT (Sendmail Ver. 8.x.x / 2 Nov. 1998); with protocol ESMTP and identifier SAA08663; on date and time Tue, 3 Apr 2001 18:06:59 +0200 (MET DST). Message-ID: <007801c0bc58$da1ef620$d046af83 [at] elet [dot] polimi.it> From: "Luciano Baresi" To: "Luciano Baresi" Subject: GT-VMT'01: call for posters Date: Tue, 3 Apr 2001 18:10:17 +0200 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0075_01C0BC69.56765F00" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk This is a multi-part message in MIME format. ------=_NextPart_000_0075_01C0BC69.56765F00 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Call for Posters Second International Workshop on Graph Transformation=20 and Visual Modeling Techniques Crete, Greece, July 12 and 13, 2001 A Satellite Event of ICALP'2001 SCOPE AND OBJECTIVES OF THE WORKSHOP The growing success of visual modeling techniques in the design and development of complex applications is justified by many reasons. For example, the use of visual representations simplifies the communication both among developers and between developers and their customers; furthermore, visual techniques provide a high-level, yet precise language which allows to express and reason about concepts at their natural level of abstraction; finally, the availability of modern computational resources (like graphical workstations) provides easier access to tools supporting such techniques. However, despite the wide-spread usage of visual modeling techniques, compared to textual languages there is a lack of well-understood (and integrated) methodologies for defining their semantics. For example, there exists no equivalent approach to operational semantics in the SOS style, nor is there a general method for defining denotational semantics. The same applies to concepts like type systems, deductive proof methods, etc. for visual modeling techniques. On the other hand there are a number of individual approaches which exploit formalisms like Graph Transformation Systems, Process Algebra, Abstract State Machines, etc. to solve certain aspects of the overall problem. The workshop aims at bringing together scientists and researchers interested in discussing formal methodologies for the definition of syntax and semantics of visual modeling techniques. In particular contributions exploring the use of the theory of Graph Grammars and Graph Transformation Systems to this aim are welcome, as well as approaches based on other formalisms. Preferably, the contributions should be methodological in nature, rather than focusing on particular technical aspects. CALL FOR POSTERS Posters should not exceed 3 pages. They should clearly state an = innovative application of GT-VMT techniques. Posters should be submitted by 15 May, = 2001. They will be lightly reviewed before the workshop; selected posters will = be presented at the workshop and collected in a technical report. The most interesting ones will be considered for inviting to submit a full (and thoroughly revised) paper for post-workshop proceedings. PROGRAM COMMITTEE Luciano Baresi (Politecnico di Milano, Italy) [co-chair] Andrea Corradini (University of Pisa, Italy) Gregor Engels (University of Paderborn, Germany) Robert France (Colorado State University, USA) Reiko Heckel (University of Paderborn, Germany) Hans-Joerg Kreowski (University of Bremen, Germany) Francesco Parisi Presicce (Universit=E0 di Roma, Italy) Mauro Pezze' (Universita' di Milano, Bicocca, = Italy)[co-chair] Gregor Rozenberg (Universiteit Leiden ,Netherlands) Gabriele Taentzer (TU Berlin, Germany) [co-chair] INVITED SPEAKERS =20 Hartmut Ehrig (TU Berlin, Germany) Mary Jean Harrold (Georgia Institute of Technology, USA) Andy Sch=FCrr (University of the German Federal Armed = Forces, Germany) Alex Wolf (University of Colorado at Boulder, USA) FURTHER INFORMATION Please contact: Luciano Baresi DEI - Politecnico di Milano Piazza L. da Vinci, 32. I-20133 - MILANO (Italy) Fax: +39 02 2399 3411 E-mail: baresi [at] elet [dot] polimi.it Mauro Pezze' DISCO - Universita' di Milano Bicocca Via Bicocca degli Arcimboldi 8 - I-20126 MILANO (Italy) Fax: +39 02 6448 7839 E-mail: pezze [at] disco [dot] unimib.it Gabriele Taentzer Technical University of Berlin Computer Science Department (FB 13) Sekr. FR 6-1 Franklinstr. 28/29, D-10587 Berlin (Germany) Fax: +49 30 314 23516 Email: gabi [at] cs [dot] tu-berlin.de ------=_NextPart_000_0075_01C0BC69.56765F00 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Call for Posters
 
Second International Workshop on Graph Transformation
and = Visual=20 Modeling Techniques
<http://www.elet.polimi.it/GT-VM= T>
 
Crete, Greece,
July 12 and 13, 2001
 
A Satellite Event of ICALP'2001
<http://www.cti.= gr/engl_v/news/sc_conf/icalp.html>
 
SCOPE AND OBJECTIVES OF THE WORKSHOP
 
The growing success of visual modeling techniques in the design=20 and
development of complex applications is justified by many = reasons. =20 For
example, the use of visual representations simplifies=20 the
communication both among developers and between developers and=20 their
customers; furthermore, visual techniques provide a high-level, = yet
precise language which allows to express and reason about = concepts=20 at
their natural level of abstraction; finally, the availability = of
modern=20 computational resources (like graphical workstations) provides
easier = access=20 to tools supporting such techniques.
However, despite the wide-spread = usage=20 of visual modeling techniques,
compared to textual languages there is = a lack=20 of well-understood (and
integrated) methodologies for defining their=20 semantics.  For example,
there exists no equivalent approach to=20 operational semantics in the
SOS style, nor is there a general method = for=20 defining denotational
semantics.  The same applies to concepts = like type=20 systems, deductive
proof methods, etc. for visual modeling = techniques.
On=20 the other hand there are a number of individual approaches = which
exploit=20 formalisms like Graph Transformation Systems, Process = Algebra,
Abstract State=20 Machines, etc. to solve certain aspects of the = overall
problem.
The=20 workshop aims at bringing together scientists and = researchers
interested in=20 discussing formal methodologies for the definition of
syntax and = semantics of=20 visual modeling techniques. In particular
contributions exploring the = use of=20 the theory of Graph Grammars and
Graph Transformation Systems to this = aim are=20 welcome, as well as
approaches based on other = formalisms.
Preferably, the=20 contributions should be methodological in nature,
rather than = focusing on=20 particular technical aspects.
 
CALL FOR POSTERS
 
Posters should not exceed 3 pages. They should clearly state an=20 innovative
application of GT-VMT techniques. Posters should be = submitted by=20 15 May, 2001.
They will be lightly reviewed before the workshop; = selected=20 posters will be
presented at the workshop and collected in a = technical=20 report. The most
interesting ones will be considered for inviting to = submit a=20 full (and
thoroughly revised) paper for post-workshop = proceedings.
 

PROGRAM COMMITTEE
 
Luciano=20 Baresi           &= nbsp;   =20 (Politecnico di Milano, Italy) [co-chair]
Andrea=20 Corradini          &nbs= p;  =20 (University of Pisa, Italy)
Gregor=20 Engels           &= nbsp;    =20 (University of Paderborn, Germany)
Robert=20 France           &= nbsp;    =20 (Colorado State University, USA)
Reiko=20 Heckel           &= nbsp;     =20 (University of Paderborn, Germany)
Hans-Joerg=20 Kreowski           = (University=20 of Bremen, Germany)
Francesco Parisi Presicce     = (Universit=E0 di Roma, Italy)
Mauro=20 Pezze'           &= nbsp;     =20 (Universita' di Milano, Bicocca, Italy)[co-chair]
Gregor=20 Rozenberg          &nbs= p;  =20 (Universiteit Leiden ,Netherlands)
Gabriele=20 Taentzer           = ; =20 (TU Berlin, Germany) [co-chair]
 
INVITED SPEAKERS
 
Hartmut=20 Ehrig           (TU = Berlin,=20 Germany)
Mary Jean Harrold       = (Georgia=20 Institute of Technology, USA)
Andy=20 Sch=FCrr           = ; =20 (University of the German Federal Armed Forces, Germany)
Alex=20 Wolf           &nb= sp;  =20 (University of Colorado at Boulder, USA)
 

FURTHER INFORMATION
 
Please contact:
 
Luciano Baresi
DEI - Politecnico di Milano
Piazza L. da = Vinci, 32.=20 I-20133 - MILANO (Italy)
Fax: +39 02 2399 3411
E-mail: baresi [at] elet [dot] polimi.it
 
Mauro Pezze'
DISCO - Universita' di Milano Bicocca
Via = Bicocca degli=20 Arcimboldi 8 - I-20126 MILANO (Italy)
Fax: +39 02 6448 = 7839
E-mail: pezze [at] disco [dot] unimib.it
 
Gabriele Taentzer
Technical University of Berlin
Computer = Science=20 Department (FB 13)
Sekr. FR 6-1
Franklinstr. 28/29, D-10587 Berlin = (Germany)
Fax: +49 30 314 23516
Email: gabi [at] cs [dot] tu-berlin.de
------=_NextPart_000_0075_01C0BC69.56765F00-- From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 9 14:21:07 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id OAA17128 for reliable_computing-outgoing; Mon, 9 Apr 2001 14:21:06 -0500 (CDT) Received: from bologna.vision.caltech.edu (IDENT:root [at] bologna [dot] vision.caltech.edu [131.215.134.19]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id OAA17123 for ; Mon, 9 Apr 2001 14:20:59 -0500 (CDT) Received: from modena.vision.caltech.edu (modena [131.215.134.17]) by bologna.vision.caltech.edu (8.9.3/8.8.7) with ESMTP id MAA16976 for ; Mon, 9 Apr 2001 12:20:53 -0700 Received: (from arrigo@localhost) by modena.vision.caltech.edu (8.9.3+Sun/8.9.3) id MAA21302; Mon, 9 Apr 2001 12:22:45 -0700 (PDT) X-Authentication-Warning: modena.vision.caltech.edu: arrigo set sender to arrigo [at] vision [dot] caltech.edu using -f To: reliable_computing [at] interval [dot] louisiana.edu Subject: global optimization package for Sun Forte C++ From: Arrigo Benedetti Date: 09 Apr 2001 12:22:45 -0700 Message-ID: Lines: 11 User-Agent: Gnus/5.0807 (Gnus v5.8.7) Emacs/20.4 MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Dear all, I am looking for global optimization code that can be compiled with Sun Forte C++ or at least ported to this environment with some effort. Best, -Arrigo -- Dr. Arrigo Benedetti e-mail: arrigo [at] vision [dot] caltech.edu Caltech, MS 136-93 phone: (626) 395-3129 Pasadena, CA 91125 fax: (626) 795-8649 From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 9 14:47:35 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id OAA17419 for reliable_computing-outgoing; Mon, 9 Apr 2001 14:47:35 -0500 (CDT) Received: from mailhost.uni-koblenz.de (mailhost.uni-koblenz.de [141.26.64.1]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id OAA17414 for ; Mon, 9 Apr 2001 14:47:25 -0500 (CDT) Received: from nostromo.uni-koblenz.de (root [at] nostromo [dot] uni-koblenz.de [141.26.66.122]) by mailhost.uni-koblenz.de (8.9.3/8.9.3) with ESMTP id VAA22264; Mon, 9 Apr 2001 21:46:40 +0200 (MET DST) Received: (from peter@localhost) by nostromo.uni-koblenz.de (8.9.3/8.9.1) id VAA00694; Mon, 9 Apr 2001 21:46:39 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Message-ID: <15058.4511.66666.652288 [at] nostromo [dot] uni-koblenz.de> Date: Mon, 9 Apr 2001 21:46:39 +0200 To: Ijcar Publicity Chair From: Ijcar Publicity Chair Subject: IJCAR 2001: Call for Participation, Student Travel Award X-Mailer: VM 6.90 under 21.1 (patch 14) "Cuyahoga Valley" XEmacs Lucid Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by interval.usl.edu id OAA17415 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk ------------------------------------------------------------ IJCAR 2001 The International Joint Conference on Automated Reasoning June 18 - 23, 2001 Siena, Italy http://www.dii.unisi.it/~ijcar/ ------------------------------------------------------------ Contents: 1) Call for Participation 1-1) About IJCAR 1-2) Technical Programme 1-3) Timetable 1-4) Registration 2) Woody Bledsoe Student Travel Award: Call for Nominations 1) Call for Participation ========================= 1-1) About IJCAR ---------------- The International Joint Conference on Automated Reasoning (IJCAR) is the fusion of three major conferences in Automated Reasoning: - CADE (The International Conference on Automated Deduction), - TABLEAUX (The International Conference on Automated Reasoning with Analytic Tableaux and Related Methods), and - FTP (The International Workshop on First-Order Theorem Proving). These three events will join for the first time at the IJCAR conference in Siena in June 2001. Conference chair: Fabio Massacci, Univ. di Siena, Italy Email: ijcar-cch [at] dii [dot] unisi.it Programme Co-chairs: Rajeev Goré, ARP-ANU, Australia, Alexander Leitsch, TU-Wien, Austria, Tobias Nipkow, TU-München, Germany Email: ijcar-pch [at] dii [dot] unisi.it Further organizers and the programme committee members are listed at the IJCAR web page. 1-2) Technical Programme ------------------------ The technical programme consists of - invited talks by Neil Jones, DIKU, University of Copenhagen, Denmark, Larry Paulson, University of Cambridge, England, Helmut Schwichtenberg, Ludwig-Maximilians-Universität München, Germany, Andrei Voronkov, University of Manchester, United Kingdom, and Doron Zeilberger, Temple University, Philadelphia, USA, - presumable 37 presentations of research papers from the areas Description, Modal and Temporal Logics, Saturation Based Theorem Proving and Applications, Logic Programming and Nonmonotonic Reasoning, Propositional Satisfiability and Quantified Boolean Logic, Logical Frameworks, Higher-Order Logic, Interactive Theorem Proving, Data Structures and Semantic Guidance, Equational Theorem Proving and Term Rewriting, Tableau, Sequent and Natural Deduction Calculi and Proof Theory, Automata, Specification, Verification and Logics of Programs, Nonclassical Logics, (the IJCAR web page contains a list of accepted research papers and system descriptions), - presumable 19 presentations of system descriptions, as well as demonstrations of systems, - presentations of short papers (accepted short papers are not yet determined), - 10 workshops, titled Future directions in Automated Reasoning, Precise Modelling and Deduction for OO-Software Development, Verification, Unification (UNIF-2001), Automation of Proof by Mathematical Induction, Strategies in Automated Deduction (STRATEGIES 2001), Mechanized Reasoning about Languages with Variable Bindings (MERLIN 2001), Theory and Application of Quantified Boolean Formulas, Proof Transformations, Proof Presentations and Complexity of Proofs (PTP-01), Issues in the Design and Experimental Evaluation of Systems for Modal and Temporal Logics (submission to some workshops is still open at the time of making this text public (April 9, 2001)! ), - 4 tutorials: Computational mathematics: opportunities and challenges for computational logic, by Ursula Martin (St Andrews), Tom Kelsey (St Andrews), James Davenport (Bath), Clausal Temporal Resolution, by Clare Dixon (Liverpool), Michael Fisher (Liverpool), Constraint Handling Rules, by Thom Fruehwirth (Munich), Description Logics: Basics, Inference Algorithms, and Applications, by Ian Horrocks (Manchester), Ulrike Sattler (Aachen), - CASC (Automated Theorem Provers System Competition). Furthermore, connected to IJCAR are the workshops Calculemus-2001 and ICoS-3 (Inference in Computational Semantics). For further information please visit the IJCAR web page or contact the IJCAR publicity chair ( ijcar-publicity [at] dii [dot] unisi.it ) 1-3) Timetable -------------- Monday, June 18 : IJCAR Workshops, ICoS-3 Tutorials Tuesday, June 19 : IJCAR Workshops, IJCAR Tutorials ICoS-3 Workshop Wednesday, June 20 : IJCAR Main Conference ICoS-3 Workshop Thursday, June 21 : IJCAR Main Conference, CASC, Conference Banquet, Calculemus-2001 Friday, June 22 : IJCAR Main Conference Calculemus-2001 Saturday, June 23 : IJCAR Main Conference, Excursions (not yet determined) 1-4) Registration ----------------- IJCAR Conference and Workshops registration and hotel registration is open now. EARLY REGISTRATION AT REDUCED RATES WILL BE ACCEPTED UNTIL TUESDAY 15 MAY 2001. REGISTRATIONS WILL BE ACCEPTED UNTIL JUNE 11 2001. After this date, registration must be made ON THE CONFERENCE SITE at the registration desk. Further information, including registration forms, is available from the IJCAR web page. For questions please contact the IJCAR conference chair ( ijcar-cch [at] dii [dot] unisi.it ) or the IJCAR publicity chair ( ijcar-publicity [at] dii [dot] unisi.it ) 2) Woody Bledsoe Student Travel Award: Call for Nominations =========================================================== The Woody Bledsoe Student Travel Award was created to honor the memory of Woody Bledsoe, for his contributions to mathematics, artificial intelligence, and automated theorem proving, and for his dedication to students. The award is intended to enable selected students to attend the International Conference on Automated Deduction (CADE) or the International Joint Conference on Automated Reasoning (IJCAR), whichever is scheduled for the year, by covering much of their expenses. In 2001, IJCAR will take place from the 18th through the 23rd of June, in Siena, Italy (for further information see http://www.dii.unisi.it/~ijcar/). The winners of the award will be reimbursed (to a maximum of USD 600) for their conference registration, transportation, and accommodation expenses. Preference will be given to students who do not have alternative funding. While preference will be given to students who will play an active role in the conference, including all attached workshops and events, also students who do not expect to give presentations, including students who have just begun their research in automated deduction or are considering the field, are encouraged to apply. A nomination consists of a recommendation letter of up to 300 words from the student's supervisor. Nominations for IJCAR 2001 should be sent by e-mail to Fabio Massacci, IJCAR 2001 Conference Chair (ijcar-cch [at] dii [dot] unisi.it) and Rajeev Gore, Alexander Leitsch, Tobias Nipkow, IJCAR 2001 Program Co-Chairs (ijcar-pch [at] dii [dot] unisi.it), with copies to Ulrich Furbach, CADE Inc. President (uli [at] informatik [dot] uni-koblenz.de) and Neil V. Murray, CADE Inc. Treasurer (nvm [at] cs [dot] albany.edu). Nominations must arrive no later than the 7th of May 2001, and the winners will be notified by the 14th of May 2001. The awards will be presented at IJCAR; in case a winner does not attend, the chairs and trustees may transfer the award to another nominee or give no award. -- Peter Baumgartner peter@uni-koblenz.de Tel. +49 261 287-2777 http://www.uni-koblenz.de/~peter/ From owner-reliable_computing [at] interval [dot] usl.edu Tue Apr 10 16:10:34 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id QAA00520 for reliable_computing-outgoing; Tue, 10 Apr 2001 16:10:34 -0500 (CDT) Received: from web6201.mail.yahoo.com (web6201.mail.yahoo.com [128.11.22.112]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with SMTP id QAA00515 for ; Tue, 10 Apr 2001 16:10:30 -0500 (CDT) Message-ID: <20010410211028.13452.qmail [at] web6201 [dot] mail.yahoo.com> Received: from [144.92.164.198] by web6201.mail.yahoo.com; Tue, 10 Apr 2001 14:10:28 PDT Date: Tue, 10 Apr 2001 14:10:28 -0700 (PDT) From: Louis Rall Subject: Pascal-XSC Books To: reliable_computing [at] interval [dot] usl.edu MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk FBSoftware has copies of Pascal-XSC, Language Reference with Examples, by R. Klatte, U. Kulisch, M. Neaga, D. Ratz, and Ch. Ullrich, Springer, 1991 for $10.00 ($10.55 for Wisconsin Residents) postpaid in the U.S.A. while supplies last. Remit to FBSoftware, P.O. Box 44666, Madison, WI 53744. __________________________________________________ Do You Yahoo!? Get email at your own domain with Yahoo! Mail. http://personal.mail.yahoo.com/ From owner-reliable_computing [at] interval [dot] usl.edu Wed Apr 11 11:21:47 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id LAA01742 for reliable_computing-outgoing; Wed, 11 Apr 2001 11:21:47 -0500 (CDT) Received: from kleene.math.wisc.edu (kleene.math.wisc.edu [144.92.166.90]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id LAA01737 for ; Wed, 11 Apr 2001 11:21:39 -0500 (CDT) Received: from forelli.math.wisc.edu (forelli.math.wisc.edu [144.92.166.70]) by kleene.math.wisc.edu (8.9.3/8.9.3) with ESMTP id LAA01029; Wed, 11 Apr 2001 11:11:01 -0500 (CDT) Date: Wed, 11 Apr 2001 11:10:56 -0500 (CDT) From: Hans Schneider To: NETS -- at-net , "Hershkowitz, Danny -- Hershkowitz Daniel" , Danny Hershkowitz , E-LETTER , "na.digest" , ipnet-digest [at] math [dot] msu.edu, wim@bell-labs.com, hjt [at] eos [dot] ncsu.edu, vkm [at] eedsp [dot] gatech.edu, reliable_computing [at] interval [dot] louisiana.edu Subject: LAA vols 328, 329, Contents Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Dear Net Organizer: Please circulate the attached LAA contents over your net. Thanks hans ------------------------------------------------------------------------------ Hans Schneider hans [at] math [dot] wisc.edu. Department of Mathematics 608-262-1402 (Work) Van Vleck Hall 608-271-7252 (Home) 480 Lincoln Drive 608-263-8891 (Work FAX) University of Wisconsin-Madison 608-271-8477 (Home FAX) Madison WI 53706 USA http://www.math.wisc.edu/~hans (URL) ------------------------------------------------------------------------------ ContentsDirect from Elsevier Science ====================================== Journal: Linear Algebra and its Applications ISSN : 0024-3795 Volume : 328 Issue : 1-3 Date : 01-May-2001 pp 1-55 Condensed forms of linear control system under output feedback J. Stefanovski http://www.elsevier.nl/PII/S0024379500003359 pp 57-68 Perron-Frobenius theorem for matrices with some negative entries P. Tarazaga, M. Raydan, A. Hurman http://www.elsevier.nl/PII/S002437950000327X pp 69-94 On positivity of analytic matrix functions in polydisks V. Bolotnikov, L. Rodman http://www.elsevier.nl/PII/S0024379500003311 pp 95-119 The four-block Adamjan-Arov-Krein problem for discrete-time systems V. Ionescu, C. Oara http://www.elsevier.nl/PII/S0024379500003256 pp 121-130 Distribution results on the algebra generated by Toeplitz sequences: a finite-dimensional approach S.S. Capizzano http://www.elsevier.nl/PII/S0024379500003116 pp 131-152 Monotone matrix functions of two variables M. Singh, H.L. Vasudeva http://www.elsevier.nl/PII/S0024379500003086 pp 153-160 de Caen's inequality and bounds on the largest Laplacian eigenvalue of a graph J.-S. Li, Y.-L. Pan http://www.elsevier.nl/PII/S0024379500003074 pp 161-202 Graph theoretic methods for matrix completion problems L. Hogben http://www.elsevier.nl/PII/S0024379500002998 pp 203-222 The Hadamard core of the totally nonnegative matrices A.S. Crans, S.M. Fallat, C.R. Johnson http://www.elsevier.nl/PII/S0024379500003372 pp 223 Author index http://www.elsevier.nl/PII/S0024379501003020 ------- ContentsDirect from Elsevier Science ====================================== Journal: Linear Algebra and its Applications ISSN : 0024-3795 Volume : 329 Issue : 1-3 Date : 15-May-2001 pp 1-8 On the spectral radius of trees G.J. Ming, T.S. Wang http://www.elsevier.nl/PII/S0024379500003360 pp 9-47 On Stein's equation, Vandermonde matrices and Fisher's information matrix of time series processes. Part I: The autoregressive moving average process A. Klein, P. Spreij http://www.elsevier.nl/PII/S0024379501002312 pp 49-59 Stirling matrix via Pascal matrix G.-S. Cheon, J.-S. Kim http://www.elsevier.nl/PII/S0024379501002348 pp 61-75 Numerical ranges, Poncelet curves, invariant measures B. Mirman, V. Borovikov, L. Ladyzhensky, R. Vinograd http://www.elsevier.nl/PII/S0024379501002336 pp 77-88 Bounds for determinants of meet matrices associated with incidence functions I. Korkee, P. Haukkanen http://www.elsevier.nl/PII/S0024379501002385 pp 89-96 On upper bound for the quantum entropy W. Hebisch, R. Olkiewicz, B. Zegarlinski http://www.elsevier.nl/PII/S0024379501002440 pp 97-136 The linearization of boundary eigenvalue problems and reproducing kernel wHilbert spaces B. Curgus, A. Dijksma, T. Read http://www.elsevier.nl/PII/S0024379501002373 pp 137-156 Asymmetric algebraic Riccati equation: Ahomeomorphic parametrization of the set of solutions A. Ferrante, M. Pavon, S. Pinzoni http://www.elsevier.nl/PII/S0024379501002415 pp 157-169 n-Transitivity and the complementation property L. Livshits, G. MacDonald http://www.elsevier.nl/PII/S0024379501002464 pp 171-174 On the orbit of invariant subspaces of linear operators in finite-dimensional spaces (new proof of a Halmos's result) A. Faouzi http://www.elsevier.nl/PII/S0024379501002397 pp 175-187 Automorphisms of the Lie algebra of strictly upper triangular matrices over certain commutative rings Y. Cao http://www.elsevier.nl/PII/S0024379501002294 pp 189 Author index http://www.elsevier.nl/PII/S0024379501003111 ------- NOTES: ContentsDirect, which is automatically generated, lists the first author of each paper and the corresponding author (if different). Please note that only subscribers can access full text and abstracts through the provided URLs. Visit the journal at http://www.elsevier.nl/locate/jnlnr/07738 From owner-reliable_computing [at] interval [dot] usl.edu Wed Apr 11 23:10:09 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id XAA02572 for reliable_computing-outgoing; Wed, 11 Apr 2001 23:10:08 -0500 (CDT) Received: from maebashi-it.ac.jp (zhong01.maebashi-it.ac.jp [202.236.152.193]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id XAA02567 for ; Wed, 11 Apr 2001 23:10:03 -0500 (CDT) Received: (from zhong@localhost) by maebashi-it.ac.jp (8.9.1/8.9.1) id NAA10969 for reliable_computing [at] interval [dot] usl.edu; Thu, 12 Apr 2001 13:20:39 +0900 (JST) (envelope-from zhong) Date: Thu, 12 Apr 2001 13:20:39 +0900 (JST) From: Ning Zhong Message-Id: <200104120420.NAA10969@maebashi-it.ac.jp> To: reliable_computing [at] interval [dot] usl.edu Subject: IEEE Data Mining 2001: Call for Workshop Proposals Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk [Apologies if you receive this more than once] ---------------------------------------------------------------------- ICDM '01: The 2001 IEEE International Conference on Data Mining Sponsored by the IEEE Computer Society ---------------------------------------------------------------------- San Jose, California, USA November 29 - December 2, 2001 Home Page: http://kais.mines.edu/~xwu/icdm/icdm-01.html IEEE ICDM 2001: Call for Workshop Proposals ******************************************* As an important part of the 2001 IEEE International Conference on Data Mining (ICDM '01), the ICDM '01 workshops program will focus on new research challenges and initiatives in data mining. It will provide a venue for several topical workshops on specific data mining research and/or application problems. ICDM '01 workshops will foster the discussion of exciting research directions and works in progress through paper presentations, round table and panel discussions, and invited talks. We especially encourage workshop submissions that capitalize on the location of the conference in Silicon Valley by targeting an audience from both industry and academia. Workshop chairs define the focus and structure of each workshop. Responsibilities include (1) writing the call for papers and publicizing it, (2) selecting the workshop organizing and program committees, and (3) deciding the workshop program content. Submission Instructions ======================= June 15, 2001: Workshop proposals are due. Workshop proposals should be sent via email to the ICDM '01 Workshops Chair, Johannes Gehrke (johannes [at] cs [dot] cornell.edu). A workshop proposal should include the following information: (1) Workshop title (2) Workshop organizers with full contact information (3) Description of the workshop including objectives, content, and format of the workshop (4) List of potential attendees (5) List of potential authors of workshop contributions June 29, 2001: Notification of workshop proposal acceptance. Suggested Timeline for Workshop Chairs ====================================== July 13, 2001: Workshop call for papers has been sent out. September 14, 2001: Workshop paper submissions due. October 12, 2001: Workshop paper acceptance notices. October 26, 2001: Camera-ready version of workshop papers due. November 29, 2001: Workshop takes place. From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 15 05:58:49 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id FAA06642 for reliable_computing-outgoing; Sun, 15 Apr 2001 05:58:49 -0500 (CDT) Received: from relay.kfupm.edu.sa (relay.kfupm.edu.sa [212.26.1.21]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id FAA06637 for ; Sun, 15 Apr 2001 05:58:29 -0500 (CDT) Received: from relayin.kfupm.edu.sa (relayin.kfupm.edu.sa [212.26.2.26]) by relay.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id NAA08516; Sun, 15 Apr 2001 13:44:12 +0300 Received: from iws20.dpc.kfupm.edu.sa (iws20.dpc.kfupm.edu.sa [196.15.32.20]) by relayin.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id NAA09948; Sun, 15 Apr 2001 13:42:30 +0300 Received: from itcH70.itc.kfupm.edu.sa (itc.itc.kfupm.edu.sa [196.15.32.8]) by iws20.dpc.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id NAA45358; Sun, 15 Apr 2001 13:35:49 +0400 Received: from smtp (webnews.kfupm.edu.sa [196.15.32.251] (may be forged)) by itcH70.itc.kfupm.edu.sa (AIX4.3/8.9.3/8.9.2) with SMTP id NAA37732; Sun, 15 Apr 2001 13:51:36 +0300 Received: from 196.1.65.140 by smtp (InterScan E-Mail VirusWall NT); Sun, 15 Apr 2001 13:53:17 +0300 Message-ID: <000d01c0c59a$51a83ef0$8c4101c4@icssarfraz> From: "Dr. Muhammad Sarfraz" To: "Dr. Muhammad Sarfraz" Subject: GDAM_CGIM'2001 Date: Sun, 15 Apr 2001 13:53:00 +0300 MIME-Version: 1.0 Content-Type: multipart/related; type="multipart/alternative"; boundary="----=_NextPart_000_0007_01C0C5B3.62DC5A80" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.2919.6700 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk This is a multi-part message in MIME format. ------=_NextPart_000_0007_01C0C5B3.62DC5A80 Content-Type: multipart/alternative; boundary="----=_NextPart_001_0008_01C0C5B3.62DC5A80" ------=_NextPart_001_0008_01C0C5B3.62DC5A80 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Special Session on=20 GEOMETRIC DESIGN & Modeling The International Conference on COMPUTER GRAPHICS AND IMAGING=20 (CGIM 2001) AUGUST 13-16, 2001 Honolulu, Hawaii, USA. http://www.iasted.com/conferences/2001/hawaii/cgim.htm Aims and Scope Computer Aided (CA) techniques for Geometric Design (GD) are extremely = useful tools in a number of industrial settings. Specifically, the = application of geometry to current problems of designing and = manufacturing have grown at a large scale. In addition to its critical = importance in the traditional fields of automobile and aircraft = manufacturing, shipbuilding, and general product design, more recently, = the CAGD methods have also proven to be indispensable in a variety of = modern industries, including computer vision, robotics, medical imaging, = visualization, and even media. The purpose of the Geometric Design session in this conference = (CGIM'2001) is to focus on interdisciplinary methods and affiliate = research done among various disciplines with the goal of stimulating = views and providing a forum where researchers and practitioners can = discuss the latest developments linked to Computer Graphics and Imaging. = Papers presenting original research with the theme of "Geometric Design = & Modeling" are being sought. Suggested topics include (but are not = limited to):=20 *=20 * Curve and Surface Design * Visualization * Computer Graphics * Geometric Modeling=20 * Geometric Algorithms=20 * Manufacturing Geometry * Computational Geometry * Reverse Engineering * Solid Modeling * Computer Aided Design (CAD) * Computer Aided Manufacturing (CAM)=20 * Font Design * Multiresolution Methods * Geometric Transformations * Mesh Processing * Shape Abstraction and Modeling=20 * Shape Analysis and Description * Industrial Applications=20 All Submissions should be addressed to: Chair: GDAM-CGIM'2001 Dr. M. Sarfraz Department of Information and Computer Science, King Fahd University of Petroleum and Minerals, KFUPM # 1510, Dhahran 31261. Saudi Arabia. Tel: +966-03-860-2763 Fax: +966-03-860-2174 Email: sarfraz [at] kfupm [dot] edu.sa ------=_NextPart_001_0008_01C0C5B3.62DC5A80 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable *******************************

Special Session on

GEOMETRIC DESIGN & Modeling

The International Conference on COMPUTER = GRAPHICS AND IMAGING

(CGIM 2001)

AUGUST 13-16, = 2001

Honolulu, Hawaii, = USA.

http://www.iasted.com/conferences/2001/hawai= i/cgim.htm

 

Aims and = Scope

Computer Aided (CA) techniques = for Geometric Design (GD) are extremely useful tools in a number of = industrial settings. Specifically, the application of geometry to current problems = of designing and manufacturing have grown at a large scale. In addition to = its critical importance in the traditional fields of automobile and aircraft manufacturing, shipbuilding, and general product design, more recently, = the CAGD methods have also proven to be indispensable in a variety of modern industries, including computer vision, robotics, medical imaging, visualization, and even media.

 

The purpose of the Geometric = Design session in this conference (CGIM’2001) is to focus on = interdisciplinary methods and affiliate research done among various disciplines with the goal of = stimulating views and providing a forum where researchers and practitioners can = discuss the latest developments linked to Computer Graphics and Imaging. =

 

Papers presenting original = research with the theme of "Geometric Design & Modeling" are being = sought. Suggested topics include (but are not limited to): =

 


=DC    Curve and Surface Design

=DC    Visualization=

=DC    Computer Graphics

=DC    Geometric Modeling

=DC    Geometric Algorithms

=DC    Manufacturing Geometry

=DC    Computational Geometry

=DC    Reverse Engineering

=DC    Solid Modeling

=DC    Computer Aided Design (CAD)

=DC    Computer Aided Manufacturing (CAM)

=DC    Font Design

=DC    Multiresolution Methods

=DC    Geometric Transformations

=DC    Mesh Processing

=DC    Shape Abstraction and Modeling

=DC    Shape Analysis and Description

=DC    Industrial Applications


 

All Submissions should be addressed = to:

Chair: GDAM-CGIM’2001

Dr. M. Sarfraz

Departmen= t of Information and Computer Science,

King = Fahd University of Petroleum and Minerals,

KFUPM # = 1510, Dhahran 31261.

Saudi = Arabia.

Tel:     = +966-03-860-2763

Fax:     = +966-03-860-2174

Email:    sarfraz [at] kfupm [dot] edu.sa

------=_NextPart_001_0008_01C0C5B3.62DC5A80-- ------=_NextPart_000_0007_01C0C5B3.62DC5A80 Content-Type: image/png; name="image001.png" Content-Transfer-Encoding: base64 Content-ID: iVBORw0KGgoAAAANSUhEUgAAAlgAAABNCAIAAAD1gI2KAAAAAXNSR0IArs4c6QAAAAlwSFlzAAAO xAAADsQBlSsOGwAAVUNJREFUeF7tnQmcXlV5/2fJ7FkmyWSyDtkTQkASICxlURHBtdYdSxEUXKit /q3Lh49WpVr78S9V/2otloqyFHGptVRkK4IKlUiAJEDIQkIyJJBksk2Smcxkkpn5f8/9ve8zZ869 7zvvzDsTlt77uZnPveee5TnPfXN+93nOs5T29Dxe8iIc9f6YK1fNfK7jyO71a9aUlHY9mo+aytNK Fpf0NixcfFxNxdIlzV7VtpKSoy/CPNIhUw6kHEg58OJxoKW18sCBirZ91WtfGNvWNZx0jK7s6622 voYlV/es1XZtJbrwy2c2lOahpnl37xBoZehgFHWSf6xCBio95kCYgUDAD/puWbUa5Ptjd4bUc89w OKeb0p6lvWUr/QuuVz5aapXPKC8BFy9dcnL54tJXVWzJzra1kGmndVIOpBxIOfAy5UBHV+nzO2oB v32H2zdunzqisxAcGhAKigyN7NYvDOiJo9RwoWDx+GekHksg7IPAx1etMkgDz5ae1gvsiSbA79Xn j/7d/W1IhwaKlKvQrxb0cMqSJZ6MmMLhiP7vSDtPOZBy4FhzAOHvheYxgN+OPVMHFP4AMNDrUGvH gDULmYZ6S6yJFJiIjrmkQx8FfRwNpEzGCkTPQBwcRhRkrGMDhH0QeMMPVv/4AcfPaz+TAT+QD0hD tuNAOgQUl85wHx0rt1UI9qhgJdS0l2Hw+fjjq9Tnn7+25IorT87CYYqFhfzC0zopB1IOvHQ5YMJf 4ZpPX4YDPMBCTU+I6Cs8C8RI6zAPmzRKAJZxbAt6CARKnvpNAnxVW/APKB1eFDw2QOhQEEWoQeA/ fmrJxEVd7AgK1bTtZ5in2SIOBizzVaY8EnyqDoiIOIiU+ZlrXQkQe8Xl47PNW1+6v/GUspQDKQdS DiRxQPi3dUd3IcJf0IFw6/ipY63ch0O/cgCEiRhZOApaz3HZMRER4yhID/5OpDosfN9xyD8lkHWk JcJ6iPvhD2d9+huruAACl57SCyKyzyeNKGYvhoiGalyYCtTmpv3CQCIMmiMOGhz+5tZUNBzyDyNt mHIg5cCLwIGjPUe3bBtXjOVLsKXnz0HAYwIi1z4Q0vDQzmefP7ST8qnTz1I/VjK9dnLt5DmBycyA DBKGxS1c4hAIVT585hIlh10QhDypakcOCEchhTPAX/1VK3pLpDSENgmFKDBff8kJe9ZWotI08APS KOFWgCeJUBak0ppKIhRABgLlqxdmeqPzyz/jzGc0kCcatg74ztIKKQdSDqQceLE4oP2/wvWfuejU Zl4uIAmw0IBQKPjgY/fs3fxUdXXV7JPOrJvUxBDPrl11aPszGqt26vw5i5acvPT8xM08oyfRuNR2 +xKlwES1Kh0GAw0vCgYGOyMEhPVM44kjs16zxEEdwtnKx0sRCrWHR4mASqDFbWA74ytLxV/QUZ4V MhmVNAmyWlfW8+suWU3hj35wAn8/cOXT/N27xhETHa32ttKLlAMpB1IOvOgcMBXoMBp/5sLCRAUp WGgoaJjX2XkYOKyprek41MG1cckwEgHRChEi23dttVsQdP7sswLFbC4+++IpdfJsMQ4vCooeHwtH Aggd8LApKEz67aol13/SGbMgn739gqVf/Jq7FgTyYv771qe5nXteCY4TlJjZp5wr/MN/hNPFg38s 2fT7jHUM1UzW/Kd/qr/hxn1sFkoGFQ2pmvRF/w+fEpByIOWAz4HhEgHjXDULz8DwJEAda2iyICVA ncp9LOQWUOSvcNEwUuUUqr4a8hTZcXLDRBARvETVp/JA+WmFAf2JWCgUHAkbGcPCYQfCeigWCiKl AUtSjQJFlBsscf3hf1ltSCaQo5WJhjwKDsDSBEGrb/gnQdPv3wiYsLiVRykWxv/DpCUpB1IOHHsO bHyuDiuYYRQB/SkE4qDvgbBu+4H4ZH0U5GnX4cNdXV2VHFVVAkWDQLuwTg7sdx1aTbBQzUePGaO2 0qYiIBq8GRjLtSOR+bmwMD8QCtIGJTiOnEQYoqBACOWk4BBjmQ9+cIuuzdXBNygt/EcZby7cNXEQ w1FGD8A49awonMNpzZQDKQeGkQMjoQWNkwcQ5lJL+iBEQ6wU0WpuWn7Hzu3OQGbsuLEA24GD+931 mHHqmULwjwqCt/ET6k1ZCtRRn3Kq8UgQ2Hm4A5WqcJRyyY5T5p68YOZp2No4aMzin39BeS4dqYTa AuFtsFg4QkDYDwUR0fILhWASECi5cMgHYuL1H3EKVdPEmugpCMyBhQ6e0yPlQMqBlAPHgAOCwBVr Ggv02xsySfHdwbilqGCytWv93Xc9uGPTaoBt967dlDRMakCkazt4UMAmGgzthHAS9ST58VQoCOrx FwTt7OiorqmhmtpSLhxlCP6WjpuBvlSIKJnPt5Gx6wCtAcJCUHDAUDW5OhkJ1Wi9aUQTUdC35Ayk QO3nsfN347UD/wZAuC9fveSX962U1yCHwaFvoeqrRg0UqZy1nUmxcGBWpzVSDqQcKIYDxwwCRWSe +C+qYLuGbEL9/v4b9u11yyDgJxlOUp0wTCKdnra27u08fLgaZWlNDcKiJD9ji5BPJYaCum1oaLRN R1en8zAypQREbk1G5NrfQTRQNGpzOdFLWTogCvpvMEDEYQfCegaz7UBhnjSi2HkitKlEG3XSXoo4 rGa+eXlfW0oAxe//sVTbh7TlApzT34+e0auGhnx+V3KWSARjyYUf/tYSrFgFitHg7keQHikHUg6k HBh2DsgjcM3zM1taknfChn3E/J7vvuGMxMHNTy73N/mk55R4Z9uEXAOEPuyJbIS/zEUkGoKRNh3A 0ioIOA1TJU26+p7KlFsfEa2fuJOGj2EGYIUAoV9nRIHQ4YowCaiTSzsXBn4+CgosOfxwaMuuapWC NLCpsYbSoOIUAXdkEaPKUopazBoBJI8kDkowNfNRu6Cft531QtRH67D/HNMOUw6kHPhfzoERNYdJ 5K0FhUmMCOqLVs/t2Xrbf969dsXvhHAAFQpMX9tJoZScM2Y0SVlq0p6PiNRBTOzDv6oq/1boaHJk dOs0qyZ0UqJNR1SmXJuVqT87OchxGHr5+Ee53QZ2QFgABVyac0JNfeVCt+J3rR9fVccFIVuffbrD fD+GxWo0oxQ1pInDIU4U3Wt6bUfQE8sycqQBm0BR4p1sbWi74taMIhQp07YDrQkXhq8qFBbKSYML BXXDrdCH58hwJgXC/+VLVjr9lAPDyYGRc4rIQ2UhKCgsEQoiC7IviMLT4VNWgLML4Rkw1tg4hQuk Q6lGB8UmyYgBUloP9fUToqdOrDTDHLM+lWzKniV+/ShRwTCqgV5Al/Y1TblKuZBs5+49vfu3cev7 cvgWsBNmnwjWWjXVlHGQpNvigbDecAiY8SUw86PngkxJV3x8ldnFmG7zb25s1b6gpDRTdWonz/d8 8AHSsFCA6kOjeG1YCATijG8SqvS0IjIbjzTFwkH9wtPKKQdSDiRw4JhtBwJ7fkQYIyXIlBQPyxLI ghLpJLclvtH6+nrEOAmIgwXC/D8RX5Waq6ZoAzIlR06e6lz4tzZvFX4D0trUNCQbYMSaGlPnJtYs K/5HDQ6h7QTJUIqCTHJjB2nomWgyLr7okmZ86oWClGvPD30msCcUpBAUJBINu4O65W/gU3/VsiUU AmmU06GgjnGBUrrimqEp5C8HBAC99AMB0EMhddCUMi4xbiBJ/RQ/97SHlAMpB1IOIAguX1X1wMoR sQsF+fwTbtttLhSMvxG0iI+t3I0sKLtQWbUkApIziolOZDXEMv6CRoDilMlTrb4qxEcpBOFoBcjp 9K+txLqlZMfO7Rs3bmhp2fHMhg2cQsGSnv0tO9Zva36Cv62trYX8/ITleeC8SCCshwgwBmzjAwR0 UdhrCgWHIBCOg2CV9gUBS+Swv32/i39GiSxfUGBKOHv0FgeQahuf22mXZsrUP1gInnEBlKpzJWCS Zz0H0Eu3BoHUoaGgkcDfVMBINaroppAeKQdSDqQcGAIHEARXrZ74m4en4yAvfBrwiAOb3yR4WkiH fnMJgnFLE/bG1q64Zds2J1HpdK4OkV7UPySEIW/Nm78AIQy3BwQvLGjkHUh5AIE9Rx2sBj0MyAFV SBRGDSMNj0WSNKgOyXqcp6OOQ9kswXmQ1cDPtLVW2b8oEgid3AbGIK6BK5LGzGSGa2CJCr7YRyHC nzBMh8KNcigGN514+XX7WEoOekmKkho5hGc6JHe6wiXNCjQKVWS9gAAukAJpC2FALJgtRISqVCjs 4296lXIg5cAgOSBBsPhI2YMcNqw+YBZAmYn+4bGnfPWgE5Kyxp/qUfuCqCIBP/n/yb9CPhUoIbnI bCVmpauyURmPQyoAira/GJ8RT+OoWcjEGTET6aYKJ8WqkrJxnUeqdWp0H+oSZT4fJnONWCwQZuSz UxyunHJKPzgEbIAlDGQY29lwZnIEOs76GEYFAEmAymEiXZxiSYpOvRnVV886sh26zgFapEwO5X7i WgpSIBBExAcDKFVX3BbyJtI6KQdSDqQcCDiAaagEQdux42KkXeYT34Ic7xITO1COReVNt/znivvv EOxJ4OOMFJ7uROcpzaesY6Q7laP97t0tupVbRUYz2Z8IIVxb5/5DR3sEpVKl6rC6tXUNPmrKKzHX j0pPOekKZ0TsWkUDt5HPfj/FrOljA9ucXPrbxEGLAUI3SYW3BlGcJHdKL7GwgUOhI8iHTCZtZxbe WqWKVAUdVECJKlQDq0wcZJMvoJhHtgVIEzkXcnJomzDqnCFKLr/EwZ4GsgQXQCBCp0EgXV23QmQ4 ktIj5UDKgZQDhXBA6tCh7QjGVZ2BLlRo6p8BSYlYCxZygoXAHn91IgiyAN5z53UbVtyPzYsg0G37 RbiC5OeEv8pK/uIp4WxSuroUYo0Dx4msmYyDTycLVjlYMmIAuXnzFvAX/APhRlePmzFl5oymmeY1 CIDRrbBQKGu7jBRyG+0+Yo/j+SBGCBcB6ASGkz5WxCiWDUTSrdPQZuHQ9KthPzU1VIMef2szz8st BggzqkVyBAIw5EUCDl1c7FOcElJqTCENMpmv7USYM6OYjfc4SxYO4SV/2VCUwIe7hdHN7e0PT7O9 RpXT7Yrr6uU46PSffcYvbehRgT3KNRA1TQrklkwXDhFPWSLhspCfflon5UDKgZQDcEDq0BVb+vK/ H2O25Nk1lGgoUFy98n4EwXvu+KfnnnjYmb00NErBCKjIj955F8hTIgqQrWvn/15ZyTWCoIOirNDm 2kaYRIVIUKunQ7wA+Ttv5vHHNc0EFIM4MnTCWNTnqWK28VfConnZ+5Y4kWDqoHr+ggWG0Ean1LOZ 2wiVRZsQ0QFqdKE6Eg0FyTaoLyDGhcVi3CdchGtSxiPtIc8pfAxwCC6iigxc4z2nvXogTZkCLSOE lRT+k/Kc4jNOhNkgNa1RJ/X8k/eFueFzQdYnUQgQimx2NzHnSR0KC+d8WjPlwP9aDoCCazaUFZM4 YrDGL4NltXLKk1AX/JO/vPBAGs744QdX88PKWBw1iyCqpxZNO96VMjTFA5YqVLfQl0PyqB/IRt1m kDh6akmdArKNQnVlnv4OvKPAb+YTIv8K6pgLf2JYADM6HbJE6JAGFCSJrgRBHWSfAmNk8CKzzP7G L6StL/ndeoeCfrnZOCFHgmfSagYHhTyVoMmRbdLKNdjGX7Sy0eEIE7CpsgO/SKfKuE5sfbQUmVUW NJAtC530SDmQciDlQH4OgILaFBwyo0YOBV0yh53PrnrkZ4iAv/2p04VCpJMCI4ARnDicONwhk1Gh QiIKuvKqKhdfOxs+G3ShCUjDwa6hpm+Spa4VuVupmuyQrOln91UYbquQCW0aEcmgwkgOX0j1OzS5 UMSLwoz6VMFOI2kVjShdZTqJRElhZB9hkd41UvY63SznkCXCjBu7PNZJH39izxKJWciFaB0Rs+QC r9RLWZGrHmdBedZ75fVsJSJEAnWoOo1Wc5P3hT+eWmTtyPqmFeTzU0/4oqdfrkyHIk9/BZ+oTLOD 0tXL/iDC4aNbN67p3PXLTY/7k3n73FMum3/WqLJRbT2HfvrMY3q0uHrSmTOPzz/nzdtfuP/ApqA3 mtDh+WPnzp46LVdzNcz19L3zTx1dVmtPlzevg+ZclRmoZnT1r19YM7TXo2nmpyd/z2+etnjKmAk3 rH+wcAIY9MSm4/w5WtsRpSR4p/7rTiT+ioXn5n+D8R+SvXf/rdkLLX52Rk/wY/7bxRcF7zH/L7Dw l1VITUxjik8fMRJACASSUGlD86OEXMHf3NJEMCk/nhm3vqkLSkhlh7AkghKqqCZccc37Q0hGFZnN siQBTuFgJN41zWyiRDmbuFCqChMEJZ+ZdGhxTU1A9DMgSoL0Yc8P6u2T6sdH9VNkaAoOoQ25+yfH yFSI4BOMLAoIUS3iFCh9I935alJu+2eHb40GTgStTBLBbNy1TE1q98+pmym3dIZR7GwKHbgSTZv6 Aej6PXBtEAhOG3izi6lETi937Sir0t+vueee55/UC048Hrzw0+DNR39/k55ymwsIdxzcC/CwCObv 8KLpJ/3g3EsBifhwb7nz23nabnjLlw1EGavpJ5/NQ/bWi79+5YO35KckT/Pvn3cZy31+evI051HH Fd/n8+Lce/8xf7X4Uw0dlI8oJbyRO970CRsR8LbXXSB5hbx6vXd7a/6gRc6ObzXoBAJveubhOOXw 0/9N8sNI/O0N9jUNWF+y4HBZhA4KDhl0wPqYw0gRapmSmJHihyk0jCYoJad28vzM8haKjDo+9hhb TPFIiR8RzeebxC+zatEjP6OFpTOU7GgJLgytLZ2vYbP1r6m5/rMXPm02iqDdp0poGs+VYVFPLc7c kFWjzloX1aJUlOy6gYJcqBCDT9ny+lagos9sfH2jULVFy+ofucxYBLre4UxjZHEDvHnlrVyrXEaq xFfT6FSD4K5HSwhA2r+rl+Udix2rD8v0gFBBHVtcWLwSUZB1kN5Y46g5YIdUoCZNAsZBUv62O7v6 kmUDcnn4ztq3ZW/LgJTk6oFpAkUD0pOfAFZnPjKG8OOAh3yg+A2PJSW8lzwoCFWIcT5thb96vXdr i6ym6+JnRydwrOaGjyZS7v8m+WEcMxR8eO284UJBMSqPUWi/pS3v+iT3QdShezc/BQZgoolFaB8g ZeU561CuCELBSWMq+QsmoTVNBL8g44TdyqzUT9LErQ2q3pwPfnQGIOfrSGklnS1RY+jc6TkjwkSt 0MssdFQoGuKpMGyCRpVtMVo0AKuTQb7qjBrWogoMHQj3rO0LoiBEAaJUCDIhrXNhe4fR1h1niSqA T9SJiHOFOhaXCOEyNWNu9ZlyMMyaRJVHa2j+oqH1enA9m6gK7ooYt6kZ4S4HKE5hLv8bb5SX6KVW rkIQKz6B125zsfv8g89wFjIWuMGiThzJ4qrUXBxkCvmHQ9c3NBDSiEyTeeXHg/xvF/1bMeu7r/I9 xpTk/8LgE8HX3A7t1Yt1syY08rf42dEJZBQoefPDOAb/LfGUwDpmeFMpBZga95ewEiaYSxxkR5A1 FhQkahoQ4ktCgpMAMCQsopkEJHCTf27nfmApY/+StQj1VaO2/WZ6SOO22yqMLE6VyJByBC8wlRK5 ZFCip0JNS+dLNTN+sbGcR0e0lwlhqFWdDBeLfSq5Vrt6tnPZR0+URlFQqt1KNzQCcWSeE+x3ZgIC RNmg/H3QoQOh6BCKmFxoMlkWkzLO7/KI4HyqbJUaWokv+fmFVo5Vi1+uIRDv/EINh5AX9GBkCHeB YcGtLwv6iH4M/msN1xC5Vq7yh1t7b2zmLLtmHX+5jY9I4amzXFISOwCkP7v7e4mAYR3SWyLxIBl7 UfaI6/v+/cFcQ6uawQPSnqj1e/ZHnFw5Vr3FR/erqUIwKBWaRk3e3X5ANePEWyv/wq9G+ZT68a2/ ez6RAGrmapjIqBGlRCP6XxWzf9MVfwtGME/VhPelz6nE34l+SPotJU7q7B+VSzIrfnbIgj4Zemsa PRiaR0LfkT7Wr51QjHVM8eTFJVEZxTz42D1k1iWbkkk/AICWfhAOLDHYSEQO5yYYgYG/I5ixoIli qmWOrENCgECKMqMYNLJzkVoSoAWNMrGwY0ntDfwyLoyR1AgNQk028+hWPYtvufS0AVdVzfY7oYT+ uTUBUdM0hggyg06Gvkco34nXX+L2CM1GRjhHrl1LK1H8T6H4HrT7CElY9NAbRj0B2S87D4rEvR+3 cNyz48j2PkwS6yovn9N9Vr3PRhav3/7hX6wEA4cFd3wxvtbEeysrKyu9zvEwOPztxp/98IGLr/xU rppqaJtnf/2pa7/3rdsqptb2XNNnttP9kYylzz98/rNnnrvw/DdckTgL1sdgskE/J32n6/EnbxY9 wSM6hF1dN4Z5y8rOmlh6eZ9r6dd6L/jUhy8+5aT3r1rzVPAo3kN8COr4u6EjR4n/Oo586Afc8k7n N/1pT0+P//bBlZ6H9/AUUpvX/RcAxgdQ4v6rg70/7qO533MiD++++POvudApaYqc3Ueu/NPxN3zc huPtHL15ixEQ/Ia/8twFV3/l4uJXhvw94DX/0vEX3NLmzDWP7NqEa8Sh7c+Y+4FsTLSvZteWdJ4m ft55dxsJT77e0vdJsEdWUxc+o2yDLSNQZsHMN5yR54bZiEopKnEQ/S0yayZ4W1aY83fy8qS8Dx6Z BY1vmBNsEzJisHOpCvoOEHPop1iJcKR/i8Pef/99xGHv/lh0mIiCrFws63EUhCDwzCeLJeaaaz5q JXyGx1GQ3liG4r2xMCVKBr4C8Gvf+pHr/Axsegc4WKyv+/ZPqVR6kYvtpMP6Z9m9/LMXXnPN9ynk OsByZhEnz++HCld/8gPo6774D9cHQ2QG6s8WCh14eyhID3956dtY30FB14P3SD3AIn+G3bMSQkaZ TdBwURIQmYvF3/zHW3lZPt+YjlCQ4+8++FegIPyPa8Ld59RVq6gZoCCtYHigYDhz3UShYPGzM2Pm DG/X748ToEfQwA9jgN9W0Y8xEyWIaNHdFNUBwh/nM5sfRgRc//tfcD5657/tWPcYKCghjN6d+BWp E31ElEgEQAaijxb9jG6zP7zRlW9sIvAzGa5PRozEKTpXhBoT6USJk0ojdSiIZSgI5plSVDRTAij6 PoK+xCZKJMCZm4dBYJ/tT9YIyNSzNBTIueBwu3ZL6vVRUPuXcvxw25lZHSydD10ixOchLloBM4iJ Er+IPqo08f5vQfnrg3KlmFdeX7+y7E7lEW/laErla6GYMjoSu/XLIQlzU9syfPlKhIkoKNmIJfKi 1513/jtPN7XnDTf8x60//3XwIX/hv4/99T3fFN9AwfiWjMkNSxafePaFJ5272FkxPbt554233b7h 2fWJco9JhEgYv/+5e2uXdN9qb4cOAxSRnSEL8Yq71m89uvPq0vtU2YH09HfNme32Lxtn1J9x7vG/ um051zdPeSLYSkSo/csrQ5ng/Q997+j6TIyoc3rnIPWKHn8IG+irZ74Lxan/e6Pa55f/u5XcfM7H 3vPB1/723pUt21oTe6CCKvP01u/e+eTHw9QDEihVZ7go8VkBu4LvA0dMZEsJfgdvAfmYLwMR89b3 nbn/aEfcWNfksBnTJr/tvefz6nkL1H9sy/pvf/mmbS/s9CUzKps4WOTs3nH5uRjI+O+CzmHvsjcu bH5y5/s/+LkXZh61p//3siuMq36TYbxma5DwMS+WUlQZB7c//zDCn/LNSn7SYcu3gRDYg/hFsgjf b92Uln4TAaG87P2tOHNCMDS1VvbICJASEiABCyn0/RwkUCbKo9TUU/VsUqbpZl15RJ5FVvNvTYT1 /Sissp4KeoWyotY2OCXFupKsAB3UccbKxRyCKN8aEzsmSqKtOLdvF9i8rHyc70dXITHFhF+IU4QI Y4h+Edoed+UulluUbkKH3OIxhAm6NXf5qLdV+PvLnT9u6VMME45Z22ATReMKBS9595u//P+uChz7 +FqvnzHWSV3eBs8XfvxdNWTxyoWCQOA3v/FJfezb8YnPve3s0z/yxNjtGVsj7xE+c7pjCQY8nMtd FDOPw62tORgEtbM/OI0NqpLnMzVOWFF79Q/7wRu9gZeX3NEHq+rzmmsyGjnrm2q9X9jRk9UMX3P3 140efwjVf3f7ovhi+o3rf2LK0gVzFr7jB87zASYg7gTLNOUfGn8GtBknH5ix88n+visQefaXlhh5 4kzxlFxyg/eFcc8OLMDiWMigjOUMVu/tewtXf9Lhuu6Z0ZUPOEHZP4SC06ZM+scvfxZkkieDjteU LH3HW887/rT3uO+M7KfG60afZL+QImeHs0RADJMC8s8bfTJDAMnoz1UBhEZMDyoP++3zO2o3bj8W e5A+5bKLAQIRAcm6jl+gNIrBCm5NJAVyK/FL5QYw8p2wyoZ8CVuGnmioscyfXb2ZkGcuEGYmoxEN kATG6sSidZtfBIXmGmGOENa5IaJJgSK+b+cStMuSapUNXNV5QLxJtMxILoxWzTBSdYauGo3bmyAL UigzFiCH3rnOHq1y1FOaCC80qCvUkTVsydT0Y41GzzPlMR+JNp7JEDRrd6qafeUIgvQmYjCo4Vad gN8UxnN3GUkvqYtcuAWRP/nBN2756VcS3dsv+fhbUTE5TVd0XjS6z2siblXoFF9/3MfO3COrfxig IKPU1NQgUoxa6BKDBYdvfxhYD6KYLd/SF9MhaMhibaIeo4O+8c7/euXPg0JWyU0z24BbnXqKMtCU peeduczoT7T53HdZgzXnApo7OjoQemygL3/uw4YE8WUaUpd8+Ez1IG+TQGClAqAeOKgMLyWBltIo x/5I1/5XDjpMQ0EeMaO4sS4oiEbhkbW3UtNHQfXGr+uCE0+3HxIXX7j+w/57KWZ2q653cn9wyEkj iGPwiS9exu9wRP9jIg7iOz+iQ9C5Ym03NtZY0G1pQTeu+hmCILag0ij6cpupQOWcQCf+rhu35r3O tQ94GTzIWsHwNGPh4kGL1IZxDwQqKyaLoSB4IyWqBZRBxJThjKsMckcHdaQ+NeeKzLZcFswy7hBR NigpbH148/nvnkqey6pM9dR2NOW2wV/DQqmOOaQlNr9JNbTebMShAyH44XvmSetooKKLAqNaqy0Q 5R/x7BN6GnM6HI20pxROQt/s0Vcujavv9eiYeFo/ubDf2C/Jm0TcAmPu+dF3/TUuoB2rSytBd2qL V+KyxVL44+uvxQwhvg6qExRlcfkDBaA/qA8bLNZsR+VhZ7BYx9HXR0q/H8wLdVrh9NomIFznj37x JStPtIdkkbUecPZgvn98cN1Jixe98fWv4US89oWneA8wwZrHEUX7tXFQL5ISzDv9HhBJj6vI6ULg 40ewK4zoHKcEvcL73vnG2+/K56I+ob7OuAqLApgf8uz4BLn+J7fn+pHQ7ebXVV77/c/ytcf5iSvf NdL/O7EUHV6vwUSC8QI8fupY1kn+UgEUJDrMpuV3AIGoQ+UC0QdgWetNwxXrkxVflZ0o5glMggG3 0EWQoKWf5ghtwd6eAYb1aZaolDjRzcWzrkGicrdZqxxd8NQ5/EX6Rm7xf3BoFO3e+dHUpEq1fUHh mTpXKihtB0K/maf6Np9ihaEXt747IzVdVopJDYS28fdErQcfBW1qAa+GDoTiGjpGYAxhzkAIvGEb DxWlQoYGOf8Sy+UamBUfYbfjeEwizJRr3Kzw566tZiDb+T34iZ+QSkFTepBcmKiktd/ES+QiEQ8Q tkDBOHj4NBOczOABkNPiFSypqs/yTYU8mEod9oriDLnkfX2WC/HFGrPPeBOBR57F2poM6ERI6DhV Br/tNOG4kLhoBEmhOWxk61Qn4rUREBcH8/wkgBwQBWnpY598X/BeiqfENydhoL/59CXTpyUILjJc 8mHJTFpEeVzCprf3nPPaH932pVwfQGr41ovOt98SenifD8XMjk8QRPlc7hmMwq8FzTO/zERRdXj/ hyIOjrSNjARBFitlTWJpkjsEulDNBUiTglGClwrNWTCjXYy2xFj9hU/aHbTMSmbqAmD4NimCzAzK Zi02AYy+yJyRg2Dm9GKy0LOoojmdW9wy4VbGYCcSBE3nCZ3UZyxzkxdqCvDcjLKJlkS2sNDw2xeF xQEJmgZgdiG1p41iFxYrQMzh9AFeHdqPZ+hACH6AefhLKEqLQEiwp0jWSoSUjWrt3pa0o2RBCsql L1UOJkxaiKzGKUsZDi64pRzzHMvf65vVCOT6R7Fx2xt+1G8XaztKFKU+AV1Fw8m6/DvCXspHHA9Y vG75ylfyoyAzYvPG4MFALrDQoxq9XXXmhflREP2hWbUYr2joB/iIL9bU7N7ZGedtoEF93fPHFS4O Wm8Y3eRZuwtx8aaHhrqcxoGF9GDEAIEy3EVa+ta1/XS8hfSTn5L4FwZgXzemT0TzORx8YXzttj5i Er+o0OJef+Pn8qMg/fPziH9qUF7k7LBFohO+HvJjYTyA0Uj8h2V3cETFQW0EkibpV3+455c//ibR 0dauuKW7Za2MYnTI1lH4xK1gydZxFn2JdIYxKFHNGUA9ZGTB/lFDLQQ29ft0g5EwFw+ErX7c7iNe EKReykZQs3I3SiSnaqw+WIo89+VWLyw0khSezXS2Zv8JMdp0zHjxRwpbaUr7IrpFCGlNpIx1vUXE C6QDkyKLLS5m+gauMDDwZRwyELrfLtDittxOcdtvspcBFzET1RagsKp/pkDHk1cvPMEvR7Fp4hqG plJyxg/KMRY1dMxWqOdCSQeFu9mQoU4vSm8cGIgKPqEnyJvoCk9xRL7Ej8TFC3OP/LjFpFgQ48tH rmXr63+fT4dJb3jcB4ySaZ8VBjG9sP9ksa5eWJNoBx9oUIMNJ/UZh//ABf6ch8P4OD6FiXt7QQ/j b9qdBwDiPeRy8WZcGe4iMyFQBn0WT8nFd/+rTQ22f/4a51uJrtKsZP2J++IgXxi+DjPxi+qf7/hc /o033mwema/I2WEaCt+gHyzkY2LAvc8R/d+6dUefXcMwDqSNwN1dLewCcoJ/aEFxB/TxT0u5WfbL O3Dbtq1yM3ee5llzf5Z+CYIchmq6tc0/dx1pGiVjCVMD6xsTBAWlEjp9hapxoE+jmN2r01iqIFzR hpyZxgjemI4Lwx3V9OnhVrBHoek5LSqNus2IqhIcIyQTLmpr0JxDJHe6DvsjujS6qmanbWRac9E2 ZCB0hAItls+IeD9ATjYXkpPGLKG8lz7etUJ2tPTxQBTxsnGoMI7jREGeQp0qxH2Ca8oVOFQH8b7J YsgFPSiprycj1nO74lan9mSg2adVQoBGNDh0DSPpUMLoS/xIVA8Guqn4FORoIXMDwM8qED86Xhk8 y78UJoJx0/3dGOJbb/4upjl7+ZuU/rh5FmtVi+9iyhPOzmnNozBkzfXuEvHe+Wh7PWAZlIeNiQpk mgB4if6aWOg8tOnf4o7exVMSBKLjCyNPmM0gOIv/hZEY0I69xvxJSJyJ6YO30C1mQewvBgwvfnZ8 LaGTFxbCWN5RIhaaZfLI/W9FLzosLhNB0nkUodDMLiCOgE/d/3MgUPin5XvXwS5Oqe/8yJ/cbtva vHf3Vk5LmwcgAVoSswScGYzxwpKZYpM6wgz2BX0VpTEwE4wt2u0zodOXnKgp4UwSoU+ea9I/nrUP Tm7oqio0t85aNYI6iYxmU2MqUBPyMqaqWeWnGcgIYmlLBW0HCr/lDmjxdGyCJoOCgnLet/kK7A2h feOjooBQu2tIgchVqBlNOwpoIY0BUXIiDIRCWpEOl3LwTxIeKCWQ44LkSlTQqQnIfYLyD39riaUq BPzI7kt0G6Fs/6yHbmiBKwOxUwgByIvInRpFgqDlTRy5/1fD1XNiNM48KZAY13e0YAnzhblEWM0v XLL8JTpafPWLHzf4zLVYE6IsPx9Y9f7pxk/H68SNLwIH9vwGhInCXCCbXvWJ97KHCuU6/UBx0BNX IDsPy/7BVnyycVtMRJTiKfG/MKDB3MnxjcnPW9QGPkm33pZ1p/CaYVecpxNwjh+PfoH8JfZCgIVD np0i5eo8ePYosBAXVUaBw/GIP0Fk1OH6nxX0g160yJ4FgRjC6FRvCgrKLiD4d9zkcQS8Vrkt2eCZ XNEtWWB52Z7O9ucOtbtEDbU1FdUVnZKf2Jzz9/yMWn+7y4xEJAia7EhlUEFw4guIlFuTjMpUOk/P 8NKHE2d6E2Gbvx0Y8I1BtW/nsthjO5qV5IJqUodmDGcilDIzTr+mkE/Eq9zcMIT09nHgt+JJ3/Zk pG0WfGpefk3Kh+xQr34yGZSQ2IA0JcJVMibdgl79syZlRrdMgdyTiffySxxcmfv8285yX52xLIN9 JaRbAn2l+dSRTXbfKkMbufODmjd8Z8n1n1wF8lFB+ejl8i8K+6dtCt7RS+U20ec9yLYT0Bp3ureE Naxro6+8OG75qbhciUcup3t5rFsTP/8Oi/XB+/7LMLKs7JR4mDdrePF9U33jFJXHpxCoYamDMJpL ik30/PvouoUKDmDHca+b6gO8n9YHUPQjfqkJkb1k+4OL9/YvTQ/YlcjD4ikJ0lT5Tvpf+8JP/vb+ G+Ihb0QYHFv71e/4H0zVH3xX/NXnz8YVBGALEksVMzv/B6NugdgPvPPvfr98RTw+nz/rkfufOYSY avGg2D7+YQgKtTt375EICAQ2TKjcvbeLmNeKDrNgZiPXoOCOndt7jjo3cGBv6jTnqN5+sH1Hy4Hu o53lo6opLCkbh64PTwOXRCLyZOewuNK+aYmhmmDDbEr968BsxFgqzSq31qFu6Upj+YgrTaZ/CGko MRMee+o7+Zmbh3ozZazG1e6gK4/0n7nIjlPoz1TstR1EoaDhon2C0DmPoLYoiZBepJBEzQjsATCm LOUWmcwXCgEnMYXCrLLU3V665GQiYpselcilAXPt1oxfCA1KOFPLVu9ltK+nMjY1wkhCz0gcRLmK OIgQiSKUDUW2NrUreZqzFnypH7d+51dxEvk2T7QdsHwUfhPCXZom7aH7khMWBtmC1Fz5KBJlQdSS votCIA4S+8OHKJygc3GZxTqunEzUtsmB3T/z6HITxRQ2QYMefOGYbwtf3+jvyYl4SP30370Lix7O qlHVBf5uiqckEAd9d/KKppI8PpqIgz4K5jI2yWWXG4/qrpxW/sSHPLvgB0OKD7qF2vse+h4Gt0F8 PjjvWyYXyPkhVNt3uL3AVqb8DOq7TPFoLls7Vq+8X4agnIaCVBYKgnws06y/QkQEQVAQzBMKsvXL iSWUw7/oOHjwINIhOlKuTZoxtzzbFzSjSm2Aad03aclMV3zRigq2W6axAonTtKZ6ahuBtheIkIoK VxHLMnmdDjsbGe13OgXvtq2ciVAtBWnGGDQb+UXqU4sIqlkob5RxW3jpN9QjVVNN7VCaBQ20iRWG giKP6VNSLBAKwDBjEQRygE+65dolgl/SLMQCnLSrp4Bqdhgoym4lblyjmsCn4C2Wj9CZmwojVY3R uZCylM65QLkKKiP/Cf+gjVuoilJSOL/7l/LxP/cmQxfro6+k4jrRsxvrgy/+n+tsgssfTPB/4Cmr oY+FMo5ITAvnUgH8cd/NP/wHf5ENdgeD2B95YAN4i+t4E40vZB5SyJG4txdgM/0Eu57fXfpu6zy+ lyZ5dEC7yoC8YackmEUQIs4fPf6FodB38YOPKn8XGZoVJSBQTSPl3/2nt/g/kmJmFzjF3n9gkxFG pIJAxsX949ikHtyxZ2ohPzDqxBMKShCUFlS2MGYIYysve4FPbdjqULCyEhRk1QacSMgHyIGCND/U cWTTpu1rnm7m3P6C04siDuqR/sYPX8vnm2v6FpsyQ6WtCkWPtJfqMIBGU5Bqhy8jt2X1pW76kVOg BDsAV/nuTZXap5vNBr5xLhaRw7vvw2BAJp8K11U2OKqpQCkUesE60ew/8rmhqakajDWrH+19gtZ8 bcjjXjuLLm9U1n2C6Qeq0QxfCvw1qJp0mMAMWkfkrRXX1ePtoGiipn6UgpSDcl0ogqjUoVKiYud5 xced3KbbQDXKdqC2/aQFBVPZI4x3iP6THpARIUNNqC+sNYUtjhOA5d41Q5lsNOPWQfGnmMrIRpWj Ts+jV8zTuUXM+t2KGwxsLn3vF3720AN+qoegB776cyUIpEM8F/GFkLsh0GseaX4TeqDPt889xaQH sjc8vexQXCnn6+6UBkHEJBKgbskBm8uyIxc96lPN/cMfRZSwHFsmxTgN6uE/3+DgcP6st2/5/MQ4 6zRx5JtEzgyBkoAM0fCTN3xIoXyIg3rRB/468W2iB/7uNz7jU0iij+uW35tLj5r/V6rwsxjEYgpU PJ/Zl42HObUXFEzZubde8bkBDaSL+V9mbf/19tkF9hPXiGIL8zz/otQQdGJqOlPKsUyz+LIcz5u/ gJ1CLdms8tuan0Dg88c9HGFe/PMRUKR8QkMTOXi1SWZehr7qT6ah6jCADYsLanakqiYPBPXJrWks TU1qwVl8ZWygwxQN6tBXhKrERcEW2kWGrGprkBkn1YgxDaf2Vs2wSN36pNromo7ICEJ7G59N5asS HwjruQd+fN9za5b/wiAKYPMhkFZAjqDLsJBCQR0XBm/arkOrKXlOTQDU4DrASMbCgubGWzP7heCf LEhlcQoxwC3faAaH2iwUkdi4DjYTIZrVyITHUXVsDhadudPfEk8ANODoWrnQSfooSCuAkDDcQ0BW dcjmjTnd+3s8cXr8/bY3X/Q3v5n+XBwI/cU6f2/qn7VSOJQ4/UJ6yMU3KME/PZ6FI6hvO2SJQKjK KKJBwcEmN7aBBqTE3x4GCC9404fiibHiu4N69bf94q7ELFp5fk76+sGYE42lkLVIPtNJ4T3wqyMo oMWIH/BnX2SFwoGQgcr2ZXJ4gX+EBtVGIMCm5dhHAq3LIAHlbO+BglxIKepMTtp3+EAICubRoBwu qWwY1zijaaZtntmKb7rQ/EDI0MEmopjmA4+Bh+0C2mabz2HfYsXvlms/7raaWHhSf3RDLOs2F4SL aXYYHDoGdrjYNzwSxKoOAGnKW79P7Qga06RV5m8/INTuHU6B+LwHAc8G/IURsYx9O4lxgI3SQUga E/z4WIiBDJXVpwGk6iy7qtWEQgNCZZygMmKcb2ijJn4JdQwFQU3i3SA4grJgmAijgr9DOeC8rAKz yJLdWnirImsSfaqu7mz3I4vlFMzVs2WSm3vczHsf/l6geMS84nNf/XpiBokBOwRW0YjK7T0xhaH1 ENjyAIT3tD0ZyCKssFu+fb1UXvl7s24DSw2f4ESLngKZL9goBL32XfEdiWK5gJCJI7MWmGM9Tl4h lPhfGMz6T2b/efm/nBJ0lZiuLzH1Yx4WGQRSR7IgF8XzmU7sg4MhEiOGU0ejc7Hu0Z/lN5Au8C0X Ui0PEK7dt7n+0A4Az/oB+XTtq0CFDVpb3X/baCOQC628rPsm1ihJQuvezahD1Y+Un3nEQSoAhDOm zCSleyB70dAHQo2lbjWQiYD+TpsqmObWwNsQLhFLaJLZnow26qyOiZVUCIxuAvHLF5cVks1PiKE+ fWNXOoRvpvbUhShn91GhUIWFOhSkRoZFJnfaB4ouMl8hEYhSMwRCE+moKh/5Qg4/tpm0nT4EJmKh 8ijJR8LHQm7jPViJPdUQQkHbdFQCXuGiCabK7uTPq5AZxeugdCUMDf0fS4kQMpQVlgvQi3x7eRYO 51udTaZKcBNihcTNSZAhlOfWuW2dMZ4I2okdypHLT8xLBE6cHGxJyv9RH9j4OVlk2/IACP3FukAR 4b9nfixXJJ0Ce0h89QhhuBAMiF7wZN/3bhFLR406LVG0YlIPnbWzGHEwPyX+1wNkSGEQAGFQx6Yc vPo8tqb8kLDBka+k//XDbZF8DsRB9ptJ4hj/QtIPj5/ofXf+64Cxk4b23zmxVRwIUYH6Ok9fztO1 5Vgw4w4KA8MQbQdSjsno2ua23Tvdnk55xbQ4EEr5mWs7UDRPmjIfIDTrStMc+jMy2KDQrEV8s8kB mWbIrZpyWDQzTsN7HhliMShirg+0JpsaLlLfLEvd7l2UvNAVeokyfA2qviQ0HRzzqQmf2w9u5qJu zGyzBjKNrjw6JBcCqNji1tfXo0nmVmphVNMgH7ira8o7o0+WUDUKJiERZgOPDciu5Ar0IJCjN2oo F6DJhab/5JHgylm1ZLcAZX1KZdmCStXJwS2+gNLBCvC4kJuEDmlN1blcMnDVV011zrWXDWPQUzM5 9RgDoRJ/++TK3coONu185zZWLtLo5NlTec2ffAQLdWvOclM+uZ8NZCEdIuTd9d+/TWQiHT6z9b/8 r3i3O/Xtn/qjTC4dQ5YDs4CQwjb/KyEv0lMbbsmlFw0mVfjbhZnIHDu7DiBa5W9lukGq4RCSWPkP m3989fu+5bN3eCnh+8Z3NWFjddq4C4LfA1HOE7O3s9984oJLySiZ59UHgQLiub2K5DO/ivi79n+B 9ttjUgPG0S2ctwXWvO2u2YqvRgiYXe3tiIBcKyOSZbmTABRsp6l/Ez5MHNSFNti0mjdvXGFOETwS 5gn/dBgKSkcqYxkzmRkzbkZj4xRlfqByIraZn6JJTgNO34c9wzxfJhOkBSUB5knGUmJ6nz8SziR4 QQn0+574hltSchqpigvjc1uoOX5cJjYIXxL2IrRTaAjKtRpqY5LvBkrcZ0frXmCvuqoKSdEc+TVo grEMEtWAjMtfwRJHLN659xNfe064KLnNhDbfdhTcwokCy0/fNTDXEPTw5auX/PK+lRaPxtBUcOuL gMLL4mckYrI+/q1F8mewzaXUGrCVS9X2gcvwts5vYocY8c4Lr5aUmf9AufrRz70XE9BAspRGLlfb YLGmmvSxfn2SCVg6QHo7Z+5f5HFUV0MyD+RCd5N1BppQwnMp/fLgutqwLjev+y9jbCIQMvErrniH BO4hHANSEv/CYJSAEn4D65/5j1yOJbz6Vy+7gvy6A5KHAuAzn7o0kMaK57P/rmEXe5aJ7z1QPwxI 7XBVeOCR6uXN5Ud2bdKen7rdu/kp3+DClmYrNPnD94ejocGhKeUO7H0GFOQRfhHsC+bZDsz4DkZ2 pAaNFM6dOxUA8IWzxLn7DnPBbqXqS0yUWKkSAwbr0KxarEImFmh2H85MdaggV0jVRAgDDoVbgkAT +LRFWt64yEYh24YojAw7HVUAlfsbtZJZDbeWBJhQA9ziZMlfPEy4ONAe7Qv2RMJl1tsyImMCL8vk TovOk8ixEZEIGUmylyRLw7/AFsbXalLNTF0SCbVCv5rpVwMTU20uKo5M8QKuDY2DBxY9x1gi1Ois Qdd+45b7nnokHtyLte/VZ59GZoA8DuYBS9l6/OdbbiepehwOWW1ftegEEtO/86ILcmml8ogFiYt1 AITBYj0gCEE8iYIff/LmXD+MYsQU4G3dw80DopdtkomGOBBq4vIHz/8DTnwqoM1PiS+SWicoaX0s 8b8wEgdCiLzx6/feeNvtvmiYnVTZOaef+obXvfa9f3le4rZckXzmM8Letdi14q71l37hC/aThgkk weBj4liqQ30uPX5f13duv197fqXjZnAhScg3/TBU8/WffieWfkhtAxQEAseOcdr1Awd53gdylIBz h472VJW4dX/MmDH1E2azg7h73y7kQqUZyYBoXQMgUVm2HxgAFCk3xaMkJH8PzAjLxHbJesf75pRS J/rilBA9LqJRaBtvPoIK56RpFPwgO7pwObtb/A28+QsWzD3zLdNrnVcxFkZcEHBg9UMu2lEmu2H/ +pIIfal6XO/BfV17t25tgXuamu2wwrraUWW1dZlQA77kF+U7zJjRqlVbp0PN0dXjcN+kIfZHCXuE ks+C9EmD/b+ttL0uSVM2mKfEKQM/jaJNvgJlQZ8GkwKtQ9OXUmK7m4O1C80zTTSu2SFaB8uNYazP R33zk5kveuJZz5rQWIyLFYi4uqW5c33mVzXzpMlNkxsH6y034OwYZUfrPquG9XwxNA843P+qCkHA s8JfH4hICl//1R8zmxT/BfHbIA0TJfyYT2uaN+y/vUH9GPjc/OZ3fiO3BB//4qYfttElHNIotoUW DEpX3UdeMCGGpxY1JqiJLQwlYCFAyAVSI+i47NR5ixcsXrNhDc6FlACKTU2Z9FvjK52tWeXY0JNH PvturY/6MQrxWeTWTEiIa4P69MnVq01oAwIttb1Boy6000a3wjmzg5Wpi0mHtVPnU21yw0SZ0W5t JlCqA0iNa1YwlFPibwdyayagvo5Uoh5fD8q1IjbqM0LdKuwO1SgpG0U6p4wLhw2qSbnhOjp2728B /6iD76ahJoSFQCiTEKUYDF7SkG/fekL1r57ulBkqxqUkAsTdUJt/5kfBtTw3GJeaiHG2O8gjJD8J lxCG5Quq18gXvs/KBohCXNs55hk0+5Sf2LOEv/g1KtNh/6T2g54HnchF8kUxlhk0uWmDlAMpB4bE Ab4qPvGhH+PYV103RR2YTi/QiPrRxeKKysBnDoAECI0i5TcG1fgbCIUB1dKFzm10CIQkJCAEIxun LFSf+/aPYs8MsxHzLug6sGfLvm5EWCPYty+ViabUlTLU1M6Zb3LpKzM1TQl27O1xLSi1azW0JrJN hRiQmHJdyLBT6Otv5llDzVoGLCIvU1LXVV13nHShMG1vazvRBoxFFnwuI2G3V4JtAGH01qpsRtKv +npa9S8Z0VAzpx/h0hkOb/ePzRnVd9wBp+wOjlz1AULVVEJdxZ0B9rgoJi8uVjA+AXnAGzCOU1tI iSgXisuu50VRjRZCalon5UDKgSI5gBr/Z/95Z+cR4CNyB4yWSza9MLiQqYiJOCZFSayxzHm+JUi0 1LqFWKu5HZIIEWXyAKG2CVnlpQbUBUBYUT0J2kzXxzWwRH4JbGRAwU0tWzs6HRhoUw1I1racZEQz ITGDScMM02GCrEDvtFkuyA4B4XbufgYEEuWQXVYxRR0iSsIKzd2238Q0uTnqa8A8GQS9PAVKpbdU n7YXKJWyy9nU1aWsEShCz3ntkmc370Qadpzf2vLsC63oP31OKgodzEFK3rW3Ou5WqOGiF+FsZOjZ d5zQXmbgPuHeOP9kWjKzobR5twOtXEBFNeoYTfkrW7W43QpqTHlfmJxnNPgTznVt5AX0JNb3iSyk /oQZW5L6aS2EsLROyoGUAy87DqAd/ZvP/BCyif+iFVNLufarKJfxRbDt5C/rJh6Z7GVMYGMP/DvU mVlAcjnOoyC15d7A0pSlACHArD4FLUBIVXkLcCW1IVKUKGTdb5rZJCNSQrtRKAdEcxOkXFKakarU ReCcgoODrPtLXW56ak6pObqjw8WysKfqM/EVsx1I+TMbNogMk/OAPZl9Cq05eITSdfZJZ57a1Abm /WHlNig8ZdEkHr2wZbu0vms3rnJGMQc7dh90kAZz5FLpNgWjjwm0x5MbnEp2Q3PLxo0b1LP/KcOt THj8zxR93MgeOLQa9eO/qDsvsYMK6omFppQO8cO27jyxqZ5q5qrvKzyD5nKQMBeIRP4mFrLdiKDp JzXM39b8/QccQirZrI2MqrcO2CqtkHIg5cDLlANsnX7mym+tWdcKECooKBORPpDDNIFc+zKZfAMk 6Ji4I02dgZZ62LXjGbnM11a7hdGOPq+JaJvQDq31ATOVnoIDhSHbhLIfkSmN67muoS9/U90U3xAm yBcRgKKaW3AW84Vwk83GbHNMiDJRBMYyfkgXfTeAweZzYsrJ0+ZPFbYBsU8910pNPg5eNff4yy/7 WFND3QOP3/v9m37rEyCbIMKugnZYtbjZJUmE7/mzN/EIH03J65amg/59Y1GDxuDDhVdW/qUvfcTj cvWdd3Y+uaUf2y9755SpU4B9C/lavfyR2r//zq7g3eiWtq+aPXXp0la/vkw6ebTPacVzHjz97z+U NtR1lJTu4CJf1f7PTqgqmTrNtS2wCTO6b8fOVX8YuDokQfa9D+/YurZz6rSF/fkwcPO0RsqBlAMv Lw6Mrqo5erD0vt+vamlpGVUxijgx+/ezmjkIPBodbe3tR7u7R3mHechRZljIdUVlrVNFVtWU9e4p Lz08fuJxo8pHHTiw9whZJqrrgbcjR3vAv96eo5zGpe6SyBoiexzp6S3p6aUmZ0VF5tGYMbVAYE11 JQkpW/e1sne4b38HHVZUTThy5BCnWtOk49CBnft2HWjbzwhM59Chdjt7unsOHcrU7O7uBjPa2g62 7m8l8tvhw53t7W1Mtrun+0B00Ir6QCMX3MKWo908PsosORmLHjjaD7Vxq0d79+yhsp6Wl5fThIu6 hml1VeVS4ba19/D1cPhIecuBzkeeeOqJlQ8/8MiK3bvbGRSe79m7p6S3lyz3bQcPyS6moqy0qnJs aVkVZ6RnnXi4q7erq8Nxpqq7rLT++R17oHBURYXb2S0l6oWjpLOzL1g5L44TS1FY1N1bzjVP+YsR 6WCzT9RD0O/W58yUxFMyBUZvwdXUXwxwCv/PgGA3WDsdDHAGe2irsvAD8TcbmE3zSo+UAykHXpkc WPbGhfPmOc2eHZlQlni2Oee2Ku2r6aQOMofDP3taQzaGevbtOGVrI3PHcXX7tVMoFGS3z+n3vJwS XHM6ETB7UhkLUqopQ5OsQvjLHh6aQGTB1m0HkAVlLSlq7cKIpzfsJLnN7AJGJHGtJMAIu0oILI0u lEuozcwoWzmSd50Zi5ss19EZzR0XddeV/vLUbESja3erUdQQZemjK//noSee3dzcZkYrPNqw7rGH Vq1Uwg3x0xEZ/TVZ0CxCQbKD+7dRja3RKY1j4QxMYC/TLJK0c5nR97qtQTcpMURJH/0LruHPYIHQ KTkf/GO+/wA89TcCuc6jDk3sKH//8SaDDYuaj/rcz8BCpZFKj5QDKQdewRzAjeQv3nEqS6cWcpbj zIq+G7Wbu1WJUAElpNZWwwbAT1mHxKKa6oNKLqjMSqza2tbSU20TGhxyqwo6G8b08wRQE8Eq6z5C FX2iEc2AHy4E0ULv2mbd6ew1BU4FjmDPtc5AnXK29Aw5zMvepF7zdqemMSdjjRJ9JRhz/C8G+gS0 +Eurlr29Ps5RQk2mgIGSOw8fNqyqroh0yHUNYLnmaE/ZIESMc54bE+YvPmGmeVaQstFwGrWq/ysV Pf4XQ99AVWGG+kwsNL99/z3Cest/lOd/AjkfoizzrXDVD6hmTdjVw8MBX0P8/K5b4VIv+Qc7cxb1 G2kPXIxX8IPAsYeHe0OwR+jyEZ6WLPZhqhr3XJT3vR2B/4bKj33Q7TxMTh+lHEg5MEIcwLvxC5/+ +c9+7cL+IZNZCDRSIKGSUwZBHgXil2+1H63vTvYCCPkLBMrzAXlOIdbiYWV4yuaf9sNogqwje1EK ZSqiHlj3KfQ3BX0yHHiUjRtb1xV42skhRGKfmNbPV8/jo+8RkVjNKvSD0khKNhuiIDqMPCzZWTQb Tg3om79aiSv3Img7Gtp3mNRr6CX+N04olVcl8jFfBnBPcrN7WjHFj+7m+0sE/vVUHpSxTD0N/PCe Gi8eEcZPLhEHQt93kOZB7ghKggoWhtReVlCB8ngd5XXy3m+/y9w2QZlqQdQblb4UHOpzzSgtTzmQ cmAYOaBYTtahDDJZc8uqZmLi37JjPUtzXPBS/T6wad/h+z/wyFCQa4US5UKWMmeecSKrORpCZWUy 9wkGxSJUS7yu5X2hsQwFfZlPoWeC2GMizPwZbGoBKvgyX7xywOEAC3lqDidW08xu3fSzYdgyjMr6 gci3IWreF+YmM4XIU95m6gOhOkEC5ltB3pl8H/BXAmJ3z0RzlORWtksZLmW97623walGE/WixBEN xKlAOxrwDs90+rGTpwid/ikvw0ig5MxzqEJynaE60Wc6xCsDr8G8o6cPUw6kHHjFcoAwb0RbnXPi VJ0gEMsr66xyR8yY+Spc3TOAR+iT6GSp5TRVquQY5y+RVYRaYDCnAq2u92Ntu5hqM8Yi0xBTzdCR hjRB/+lHpbHVXxeIgDozwTYjSpDMMiioSlHgFZX7HutS8/IcPDBIUGFGwRhttunIlHsK1TgK9lX2 9hE1ovvrQK6fsle2RYEkLVNSKTYRbSXhYTIayIKgWiCRy9+R1yRQ7Gx/juCuPUcyFr/6amFE95qi N9Kvw8H+kAMtpdNALmkOzFWoY0HOLMSaDcRTjEjtxJQGOxROJQSmt2Jc7G0U4Baxj+yG/kkM0sAH P8f0WwfLlrR+yoGUA68wDpxzwUmzFjY9+9R2Tqam5TXSUro1mngufvIgwMbtZhEIlJVdyKToX9Eh SFMJOCdpT08RB6UyZRRZi7Doq44aIiBKz8nTFY9tFChSH90p/nOEBT7/vAs4cSF4z5+ePW/hMoO9 DDyz6EdRyughs4EnqOsfftNisohgHw791xrfa7T6fjVDTds6dX1iX6PQ2EkHiMXJ1wOnfTGAanyC iFHCQuiEyb79C6ywV4MnBpZEkgtpiEslm4jOaqmmhr9qRW+ukyz/RUvh7hPurdx595bAS+HUd5dc tKT6+cePx8fAn934hbsop2THjvqb/yNftHu5KHDSMzVjjgrVK1d1BoNeeDaxafhM6PPoiNdRt/z1 T+iZVFX6pjdV39fSGbhPxL1E4pSfNLuEtt64ud5pWp5yIOXAy54DZaVlU6dNOPjC0bXrNhw40D5u XB3ntud3Hz3SVlfXfVz9mMZx1cdNHj9vwYmzZ85ZtmT24oXHZc7j5y8+fv7kxpnTx48rqepxIV3q ahw+9R7WXhcOAEd7RnH29nRhB4JfxOGuo9tfcEFViLaKBwVeE3gLUBM/ge7urv2dR7nFTwCeEmOd ypwTGurPevPSt5/+gZMWncj5mtNOm9Q471D7c5uecxQa983p4riZk8fXV5aU1o6tPVxXU8pZgT5y 3PjaqvaGxvmTGifPbhq1cNq0stoJPT3dlOMZMXr02LG46485PGZ0+dixUyeMx2+kHs+EWbPnTG+a AalVow5091TJlwSkqasbXVpWitsDfiP0MHo0uZAiT4asS8PR7vJRZe2jKurM/QQ6JVh391ZXVI4v KaumTnXteK57e9qOdB2B81VVOI70TmkY01MyhoZS/OJZge8EXwO0pQ6nA79ZC7sPd+zct7u8Yurp Jy+cNWNCTVV1a9thfD9OXjSrekz9ifOPW7b0+OlTp/O3pKwL/5XRdVWchQOh28z70u07A1/Ab3zs ZOddVzouQLuDo0qWzXCOd5y4Bhbu5Ad6Hd23M8IbQc4QgTDX/0KB2T2rUiB82a9T6QRSDow0B8aP GTNq7qj1jzwL/jUS0WXG2KMdPWUsypPHjxs/o7yqlhMaiMPCX0KxHOro5qytKScqyu6Dz3d2oxM8 Ul1VOW/e9MMdB/D20zaVvNnauzrLeg53drV1H+2RlSkWNKDL+DG1LPqVNXgGHJ4wsXrChDE8FgrS POM5V1He1l7Rg7TYtmdT81rOX991z6rVK1ZtWIu3gtiCqASEMDSUc77nbW8YW1c/prykcnTZrElN YMbk0eVg+eQJDVOnTVk4Z3zjxMYx9bSpVvm0iQ7pJ46uKK3oxmexruaom0tNQ21t3YwxJXNmTRxd O7q84vCoUd21tYRombho7rTGieN6y6tOmDFu1uxZUxuhvVf9cM5oml5ZuZ/KxLppbKjm5Bp8nTlj 7IR6AtYcnj6lZv6seWBVSW8pQjVeiQChw7amxgVzl9TWdfMVAg/BvwMH9gFdzGv6tAlMhOmAgsiC zIgXwRTmzZkNBM6ZPXn8+NGIvk1Tx44b42LlTKqvPPGEaRS++sw3zZ21qHP8njkTZ9fXT2+oqyvc WCbBqkWpARWHLG5+krUdda8EhaTS7RZ4eKaqmbS6fsOssUxrtjChTq6BZPBSSACduBVPaixT4OtL q6UceCVxAMOZG274D1Bwem3T84e24r3H7JT5wQ7Cu9h18IhywqngRa46m5oPY+XvG47KglQKUpSB KPeoL3DVQbBN1H0yj1QSImeqk/UspBy3QguKZq0ULxQ8CN4FkczihfH31TijvmWbrbEDv0+/Ptdq QA95+lE1fxSfNq6N/pt++quM3jiaOKrgt7ztJOv8kce2nn5qkw2kC/9Wo6jDxGMQQJhoL0pQNPVr CSVsGN92VIVKkITLhFmyKNdEHCCLB0Kl502cM3uQKRAO/LtOa6QcSDmQ5QBYyEq69ejOI1tLKppK mkZNZp0dW7aY58QG27o7E5baZ5jKD/SssYWehfjBRx5hQcciRjXBP/OjWLBw9rmnnw5EzTvubNoi p1Fh7263PXnDL25B3Pzzt1+88bn/YdHHJ/35F1qoD2QCriRpOvPchSLGDnqwa+vK75OnutUodl3M O1+9bqNRPqh+IGBNyyNiFNPnb8C3u3/zAFPWvinqUOQ/wbwOIpQSNXTRzNFwSR8EeUbni0RP9bWh b4JCgZD4old8PHT4yz9V3CpWXFcvPwS/ZjaHQ19ZXlgaokToy6NxOgd0n6BJ3CUjlQgH9eNOK6cc eCVxQKlAH9uyHhTUYj3godUc/FMAa4xCN6zfzGrux5QRFjoAmHv6hectEhBySyYfy/Bz403fY92/ 8NXvAAgfemAVLhZUwKFwyesWA8zWhMKySVP3HW4PUgNRKFJHj3cDte1zcNKza7vKuVi5rWLOCW6j bnxVHbcMzYUeBXOkyXETXVqoLeseiU8f8ihccOq7+AsZ6i2opgxFPoUnHz8PBP3xL39isEQFYxpI L+8RSzdoeT+opsCkcpMguinplP3hSItot3WTHNlKCxw/CgXCuJ5wwB8BFRDsSPBLSlu/sheY2xUn 9jwsEuGyS5Z2r0nwqU+UCPE79A1cMWGNR/FOgbCQl57WSTnwSuUAWLjirvWPbR3NBEk866+zuabc 3bKWLA34HUoKtIwTEgQdBI5ywECUL+K6TZh9ovWphdtBwtpVG1bcv2DZ+XMWLWnftXXF/Xe88LxL sID/xqJ5S8obF3FtlVnoSf6eSIxhQLwC3VoP+d9dnk5EakNPy9jFF/p4YxCrngFI/gpoOQSrHKtf WNc0pXxCmUsiMa7E6ZA39+w5sGEfAIkEjBcK0QzIU0Hi30S2Kw/w3DPfkot+kdRS594dR2N7nz0R t4UCYVw8ys8vPZV+Ek+JoLIl2qU8ceOweCCkZ3CLADHBQUiaD/5Z179tq7zx2kJm0K9ObG9y0D2k DVIOpBx4WXOA9BS//sX4Dc2P+sjRWjul/pAzm+fCZqcSsGHHusee29rc1ra1YbzLLiR/eSWjt2Pe zON9gYa1nv7VrXr4k7dfUTFp7pFdmzYtv0PpjUj7B3CCjgE/fRp4JDIKOdTQr29dJXaip5PqHKRB GH+hjb+gEaQ2VDaOriypra9Rlj0OS+1n+fuwL9m7bRaP9vY8IzkVXCRZnlVA7vz2zTc/uXo17vbL zn/Lgpmn9YyfQ7c62iINqEbZvX7Ng4/dAzcY2iYLDVzv7spYD6kcUgFF4BkwRkQW9BYKhHFdIjAj 90FLPR/f7ZMIhfdeIa/B6kinGt3ScIiq0Twjgmc8LTxtk3UVy0g1qGmllVMOpBx4JXCgo6t0+aoq SzbuT0mowLGrvd2AcO/mp0iSh5kom1uZfISRFOgfM6bMRNwhGy2JhMgISH4+gFBr+sqffg0xaOF5 71R91nGg8dD2ZyQbHZm5zAYdLHMlFUlA9PFD9Cf2xqQCoKUahciUXOzYtJpsug1nvS/oTYBUtu9Z 9QkMTZmYUZk++3Sfzz6PfJw7tPNZPjgevOvnBDU9/fXvOPfUixoWLj7U2lcfIBQoAoSUr3rkZ8iF Z19wlSEl5RrRWvnATHlp6TZVKCiyDNrLuNzGVt83L6/nBOo4uYiHYqEVbQU8hR9ZA5zWwpsMNmFF 4T1bTWx/hsXTfwhDp01SDqQceOlwoKay97Wnd75/6cQ4AoEfOoUNIBZkk8uXCNGKbJLJO3g0EyDN JqW0eaDgxuZ1XLCg0xbM47RcgAInMAb8Ewr6Umku6CqEb5AqkY5D9OdqFUdBMQGC8wyETGZiGcrJ LW0t4J9OtaJQJ9V4qlOPCO5KwG4mC0a2tHQAfnbyVNfCOXEDuAXtdCKM6jx+6lidJp4CgZy9vTOQ QTlDIIzrEulaMV/8QwFlohLen05XEsRaoxBLGbSj8fJcXIvHEc3DX3s02AQU+HIU0q3VgSqQPjvf QTVNK6ccSDnwCuTAkpP3vPuMkmWTZ/tzA/8qmldwAmObn1yO3MbGFS7pDeMaCaqC2wOWMmwNurOk S6ea41y4bZtDQTItEJwaIZK2lANR4GiwK8aKrxLJYYZbBsP+xYCspxMfxgC2QZ0SK9kaHH90E3ab GPWIquCQUKhDTZALdYJwPPUruA4rG3nETKccf+pr3nvVktPfg5znn42NNbrlQpLf/NlnwRZ/BxQn hXXbD/jno80HdK59YSwoSKsHt3ahjA1Vo3L48+eAnAcQBoWvv8TyS/T7GcRtREkBQdJ5KhF9+/HH V+V6K2hZyREIZHr4qroJdqex0Z36tHBsS5xRLsKgP4mqAX9daYWUAykHXvkckJp0eXO5pECJgMDY 7t1OoEGaAQUR6bhlV085KywZvc8dBV2z6NIEA5N1DB3S27x3fdqXPhkLrAW9gi2xwbJbu2UgB6MM uStQDQUmQ5/alDE/Wde5KL5RF+wXli8unZG16dk2uQSrRovKidCmay6AHjp/wxvPra9caBXyTPPp bc2tT//2Ty68LLEOgqO2FeNHCISDZGVr//r1g2yeq7rfbZ4+rdpwjZuf/GCywzTXtJuUAykHXuYc 2Phc3Yo1jWjzgKi9j92NkScBM0mJ0NE5JpMkNkpj68I9d7bKTLQE7Wj/nUKkRsvuhID4qvPejHyz dsXvsIiZcOobBISmdJUpTYCOxkWrTEmgv5X4SKF0oQAhSNOx8aGaeefUn/CaAt+DDGF0AKNAqWwy sUDBLwIHiQPdc3J1BbyBgvGniRb+CHNwddboRtvty0Ph7NMqNz/a9Yd7b0J8XHzBOK41VmLP9CMy hMcBEBbIh7RayoGUAykHUg70ccBEw/W//wUyHG5tuLgROE0Z25WrzwFhFCktPCJQRCnKU5LxKif7 a897PV5xcpwACGliEqdsZIRnA24N+qCocYWCcplAdEOEAghRafa+6ooC36ipMdnS8/0QAEJt+8nm BfErsNi0/tWD0FQ2LzJ4iRNgFjH2KFGqsx4wmWFeJy89P94V/ecC1NLe3sFZshTIqbRayoGUAykH Ug6kHHhZcOD/A2uTR9mtyQXnAAAAAElFTkSuQmCC ------=_NextPart_000_0007_01C0C5B3.62DC5A80 Content-Type: image/jpeg; name="image002.jpg" Content-Transfer-Encoding: base64 Content-ID: /9j/4AAQSkZJRgABAQEAYABgAAD//gAcU29mdHdhcmU6IE1pY3Jvc29mdCBPZmZpY2X/2wBDAAoH BwgHBgoICAgLCgoLDhgQDg0NDh0VFhEYIx8lJCIfIiEmKzcvJik0KSEiMEExNDk7Pj4+JS5ESUM8 SDc9Pjv/2wBDAQoLCw4NDhwQEBw7KCIoOzs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7 Ozs7Ozs7Ozs7Ozs7Ozs7Ozv/wAARCABXAqMDASIAAhEBAxEB/8QAHAABAAIDAQEBAAAAAAAAAAAA AAQGAwUHAgEI/8QASBAAAQMDAgQCBgcGAgkDBQAAAQIDBAAFEQYSITFBURNhBxQicYGRFRcjMkKx 0RZSU2KhwVWSJDM0NlRjgqLSNXOTQ2RyhOH/xAAaAQEAAwEBAQAAAAAAAAAAAAAAAQIDBAUG/8QA MhEAAgECBQMBBgYCAwAAAAAAAAECAxEEEhMhMSJBUWEFFHGBsdEjMlKhwfCR4RUzQv/dAAQAKP/a AAwDAQACEQMRAD8A7DSlK8wuKUrVX69t2aKkhPiSHiUst9z3PkK0p05VZqEeWQ3ZXMt2vUSzshch RU4vg2ygZW4fIVUZOrrnOUvwSiGyDgBHtLB/mVyHurSvvLlOLnyV+K6+MLWpRBWgE8P5Ggc8Bjca 9PONxA05MeLWcGOjwwXT5Nt8mx54z519Lh8BSpLdXl/eEc0qjZJStb7yVS1uyQTxLqlKBH5V6Q1C 8fBYbIByU8M4rU3O73C3OIS1ZWEvOJ3J9bPjvgDjkhWcVphq3UKSlRDRScKAMVGCM+7lXfDCzqK6 aSt5+xk5qJa0uSYrpVFfejjdgFC1JA/tW1h6wuUN9DcxCJbSuiRtdA8uivhVatt3mTfEU7Z0IcZO HDAUWX0+ZQOBHwrO2Y0llTsZxt+Jzd+yAAP/ADWuQ/8AySAaxqYeMumrFP8Avkupd0dNt1ziXWKm TDeDiDzHIpPYjoal1yyHLehz/XorymnxhCitWQsdEOH8QP4V8/OuiWa7M3m3plNpU2sEodaV95tY 4FJr57GYF0OqO8fodEJ5tmT6UpXnGgpSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUp QClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKVFl3GPDH2i8q/ dHOtJKv0h72WR4SfLnXTRwlWrulsUlUjEsD0lmOnc64lI8zWskahZQSGUFZ7ngKry1rcVuWoqUep Oa816lP2dTj+d3MHWb4Nk9fZjv3VBsfyioK3HX1jxFqWfM5rHWRlAW4AeXWu6FKnTXSrGbk3ySfU CWgQfa6g1GWyW8hXMVsi5tAS2M+Vai8y0WiAZTxyon2Ud6jPbdlsqfBjkS48RIVIdS2FHAyedeTd YcZaPElNpzxHGuf3O6ybg+XnFYOcJT0AqKuSG2gSkL7cKwliHfZF1SOxtP8Ajth5pxK0qHApOa+B xaXeABPLnXIReHWG8MvuNA/hSojBrGb1JLiVJkr3cyrearrrwTpvydtZkuIA3ZwOFfX5W7gTy6Cu ZWnWspvKZTpfRywo8RVnh6lgTU4DwbWeAQvnV4yhJ3DckrG+RIyrCjmsqSVK4Y21rkFJO5pYUMd+ dbGDzClHA5gVpKyVysW27H//0OlMRQkBSjnPEg1JACRgDAryVJQnirlWFySEuYHIc67OqTJ2iSK+ FQSMk4FYUy0E4x/WsilJW0SCMEdajK1yLngykgkAE+dKh+Ij99PzpWuSJTMywUpSvnDU+KUEpKic ADJrmlxnm63OTMUpSkrGxpvoGgrAH/WoKJ8kirzqKQuNYpK2xlak7AM458KoYaShBjtZSpIIT5Yw hP8AXcfjXu+yqaUXUffYwqvsR1LEdsSSwl59x3wozI4h94DBUR+4jkByzk18uEe6aElN3yWW7g/N bKFqdGS05zGPKt1py3ol6ydJSVMWWMhhsEcPEVxUr386kXdpF/8ASJDtjyfEjQWFPuNqGUqJ4DI+ Nek8QlV02rxUW5fDsvp8zLLdX79ivuRcwYsSNKEu66gc/wBJkpOVIa5qA7CtkizRpetbjZC2lEb6 LS0hI5J5YI+PGorOlkSNfTm7NKXbBCaCklr2gFngeB6YPKsUaz6iVr2VDGoAmWmKkrlBge0jP3cd PfV88Gnlnbpvve6va3G2ysiLPuu5GUALKq5OSUxb5Y3PCKt3F9I5BX73Cs0G13LVZl6qgqTBkJOG GGgAHCOYV3z51MtemY0P0jKj3NSrktxnxkOPD8XMkjka3GnU/RGt7vaEIKI72JDSRyBPPFUq4pRj J0t5Wzelm90iVF334Kqw80+w28yyG8ktrjEcG3PxNcfwnmB0zwrbaeuTlsvDTjilFmTtacKjzGPs 1kd8eyT/AC16v1vTE1fIjNja1c2PHQeiHU9R5nFQ5CPHSACnOCRwxjcAsYx2JVV3KFan6SX1/wBk JOL+B1ClRrc+ZVtjPnm40lR9+ONSa+PlFxk4vsdqd0KUpVQKUpQClKUApQnAya0dz1haLWstrfLr g/C0M499Ck6kKavN2N5Sqan0k28uAKhvJRnircD/AErZxdYW6dc2oMNLjynRncBwHvoYxxdCbtGR v6VjeeajtKdecS2hIyVKOAKrc/X9oiK2s75Kv5OA+dDSpWp0lebsWilU+P6R7a46EvRnWkn8Wd1W uPJZlR0SGHAtpY3JUORFCKVelV/I7mWlVaXr+1xZTjHhuueGopKk8jip9i1RDvzzrUZtxBaAJ39c 0IjiaMpZIy3N1SqxP15a4E12KUOuKaO0qSOGaRtd26S086GXktsJ3LUf6D30K+90L5cyuWelVD6x 7X/Af+VPrHtf8B/5VBX33D/rRb6VXIGtbfPS+tLTjbbDZWtauQHaof1j2r+A/wDKpLPF0Ek3Jblv pVQ+se1/wH/lUy362gXFboQ06hLSCta1cgBUCOLoSdlJFjpVQ+se1/wH/lT6x7X/AAH/AJUK++4f 9aLfStVZNRQb8hZiKUFt43IWMEVEvGsbdZpnqjqVuOAZVs6VJq69JQzuWxYKVUPrHtf8B/5U+se1 /wAB/wCVQZe+4f8AWi30quT9bQLcWUvNO73WwsoHNOehrzbtc2+5T2obTLqVuq2gkcKkv71RzZc2 5ZaVpb5qmBYnW2pAUtxYztRzAqDC13AnzGojEd4uOq2pzQmWJoxlkcty0UqNOuMO2s+LMfQ0n+Y8 6rUr0i2xl0oZYdfA/F90UJqYilS/PKxbqVVWPSFaHI7jjiXG1oGQ2Rndx6Gtj+0sNNi+l3UrbZJw lKvvK91CI4mjLiXqbmlVD6x7X/Af+VPrHtf8B/5VBn77h/1ot9K0lm1RDvKXltNrabZGVuOcBUCb 6QbRGdLbIck4PFSeCfgaku8TRUVJy2ZaqVXoesrfcLg1DiNuurdAOQOCff7qsNDWFSFRXg7ilKUL ilKxSJDcZlTrhwAPnUpNuyB6ddQygrcUEpHMmtDPvq3MtxfYT+91NQZ9xdnOZUdqBySKh17mGwEY dVTdnNOq3sj6pRUoqUSSepr6E5FeayJICedek9jE8Y44r0EedeuBFAEg8Ki5Nj5sFS4rSCgnhu6V G4Cq3d9WerOqjwWvEcScFxX3RWc5qK3ZaKuy4yJcW2RzJkuJQlAySTzrkeo79Kv81bqisMg4bA5A V4uN3lz3D62+p3jwT+Ee4VGaRIcTuVhKBXFOeZm8VYghMlxW1SlBPc1KXHCI6Wi8SoccDpWRas+Y HKsLm4r3oPvzWZY//9HnzqSlRG8qB5VjQCtW0Ak9q2rJQfbKElQ61lCGkBRShKcjhgcTWwNMQ4gj gR2rIh94HgpWf6itmkJCUpKdyunlRaG0HCQN5+8aAiN3m4NZDcpxIPZRFba1ar1BD9hiavbz9sBW PnUNllkq3BIwOuK8uLGTs4DlS7BOc1pqFuQ64Zznt8x0+FQndUXl5RUu4PZPZWKxKIdThVfDDS5h IwOpNLvyLI9Iv11bWHBNkAjkd5r25qi8uhQXcZBCuY3kUfaS2wlCRuTjrUZEZC0EhPIccUuweDcp hOfWn/8A5DSsXgdqUB+oqUpXjFzTaqBNkVjo4kn51T/EQJac8duN3wdV+oq+XmL67aZLAGVKQSn3 jiK50iR48fJ+zUoZI7Z4Hj5LQfnX0Ps1qVFx8P6mFTkyWy13y4ahvgtN6FvCJAK0eEF7sjgeNYIt n1H+2kuEnUQbnerhanywPtE55Yqdarn9G6liXRwBMe5t+rSenhvo4ZPvH51M1jLRpzVFs1FwU0pK mH0p4qUkjIxXdq1VV0klvHbZcrs/O6MrLLc01rsupHdT3SMxqHwZTYSXXvAB8Qe7pXhiy6jXrqTE RqDbORFSpUnwR7Sf3cUM2/y9ZNyY22zfS7RQ2t0eJwHHPvPCvca06j/byVHa1ADPRFStcgsDC05+ 6R099b5prM5Siunx8E72XFyuzta/IcsmpP21ZiHUO6b6uVCR4A9lPbFfRZdSL1oqInUQM1EfcqR4 I4J7Yrxa9QTYGr5lzvbXrLTA9Vckxkew2RwzirHo0i53W7amWNrUhfhsFR/+mnrWNerWoxcpJWUV wlu32/0Wioye3k0sm33e3autabvdxcVbVLADQTsT34V6aUkOoJGBsb49vZJ/Iisc65m632Zd2gdn +xQSev76/dWN5xx0JbYT7Tpw2c/vHagf5Eg/GrxjNqOfZ2V+3l/yE1vY6Fp8FNhhhXPwwa2NYorA jRGY45NISj5DFZa+QrSzVJSXds61wKUpWZIpSlAKHgMmlaTV10NqsLzjagHXPYR349aFKk1CDk+x VdZ6xU6ty3290oabJDryTgqPUDyqn27S2pNWJLluCYMHOPWXubnmkdq2WmbM1fr03Fkn7EfaOD98 DpXXkhiHHCUhDLLScAcgkCpi0tzysJT95br1d/C7I4Jf9J3nQjkebcLimZBfUW1qSCClWMjh8KuX o9ajohyNUyFFMZpBQznhuPU/kB7zW8lagtGorymyyLYJ8UKylfMbxnjjtx51qteS2ISIthgIQywy nxFtoGAM/dH51aUk1fuWqvDxviIbuO3pc0GptVTLu8Sd/glW1mM3+I9PeawRvRprG7sesPzGLclX FDOMqHv7VbvR5YI7m+8yAl1xKvDYSePh9z7/AP8AtXO6XWLaIa5MpwJSkcE9VHsKiMspOFoxcNet u3vv2OFMQLnanpFuupCn4y9u8cQod66npRMxWhViO6lt1QcDK1jIScnia57Olyb3eiW0BcqY7hts csn+wq+aomp05piPZ468uraDZVyO0DBPDqah85jmwzjGdTE2tHscuSDbgiO7JMpxbhT43Ldx51dt AOOC4TG2iErXHO1SuQI71rbpYxB0LGujycPSJjShkfdbJ4fPn8agW+4u21by2SQp1pTec8s9aS8n NJPD1YVJ8vdmKHZLjNujtsamJlS33lqL+32U9zjsP71s7noG5WHSqkSL0krckhb6wk5WAPZQnjw5 qJ+FW70d2T1aEu7Pow7J9lrP4Wx/cnPwArR69vnr9z9SaV9hFyCRyUrr8v1qczR2vLQw7qSXVL+S rsWp+7n1GNLbiKUnJfdPBIH96i3zStw0c1GnTbobih9RCGWUnjw5nyqRb7TqefeWI6bMtqI64B47 g4BHU8+1dyEKOWWm3GGnPDSEjcgHFWV4bM0weFapuFWPP+TkzljmD0eqmqkeouT3EKUhYyrwwchI 7EnGa1DFsfuyvUmJbcRS0n7dzknhVo19eBLuYgtKAYic8ct3X5VTFQtVOScMWJwMKV7LrnAY7njw qFeT27HLUi6tZRpRvGB9vOlZmk4rFzm3sTmXF7UNMpOVn58qsbFimSPR8/c23/UVTdpysEq8IcgB nmePwxV/sLlknW9iA0Y0xcVoBWWeAPU8arXpBu6FOtWiMUhtgAuBPIHoPgKOffudlbRp09dJXtZW 4uUyLAdnLRBYkIZWtO0PO8kjuaiixSrJPWh28t3Bso/AOAPxr4sXVm4NiRBUzCeSSy8of6zAHLjy 41ubNpGRq2YGkzTFiMYVI2j21g8gO3I8ahXXS+55lOE9sO0k5b3Ld6MY26LMuA4oWsNIPfHE/mKp d6YkN6juLsmYmSt13PsDCUDsK6hdHYekdK+BCQGkto8JhA5lR6+/ma5DKTdZIUq2QHJz+7KwkfdB 6moW7sjqxUMsIYSluya16P7rfFIkR9QR4xdTkR8ElIHfjUXSVvN3v6bClbshLSj48kggFI547Z5C rJ6P3XLRJkS9TNNwFrRsYaWklSu58u3xq/3OVAsFpfuLEZhtZR7GxsJKyeQOPOruVlZnVCjSdK1V JZebfycn1THftuobjJmSxIOchtsYS2non31n0y6BqC3OgkAvIPwzWxtFocvNvvN2lguBMd0IJ/Gs pOT8Pzqu2iUY3qkrbuLe1eAcZqj3Vzy66d44iW2Z/sbfWkWX+2kt96Wl7xAlLTDY4NjGAD3JGD8a l2zTEzRktWobzOQ+4Uf6PERkBKz348cVM0bb3b9qJ27zBvQwsrOeRcPEfKoOtLx9K3x1DaiWIxLS R5jgr+uanM7HVKooU5Ylrqk9vQ091ul2vtxShttcya+fsmU/dSP7CpB9EurZkcvv3lhh7GQygHA8 s10LQ9gjWy1JnEJclyxuW5zwnokeX9ya2N/1FEsUUrcUHHz9xoHiff2FFLLwddChTpU9Wru3u2zh 1mdfmXFqxLJXci4WlbU5AI5n4DjXQPSDanYlktEJiYlmNFWCpvGVuqA79uJqzachWeWgXyPZ24kl zO5wjie5B7VQ9X3g3e9rLeVtMnw2kj8Xu8yaOSv0mFdUsPRbp7ufBp2rHJvwXGj3BqBtG4uudePI Vr7zYpWi34zk+6KuJkpJbbZSeA7ms0OBqZy4NJl2gw4ZWPFfeHBCOpPHjw6V2L6T049bVzSmO+zF ASVFkHHYDIq18m0icPQUabp1kl89znUtblrszNtBU2/JAflIzxTkeykj3YzVal2eWIjt6RdQosOD MFKDuKe4PX5Vtdbxrmu7yNV2lDky2SMB1JTxbIGOXbhzrHanFyX4i2cpU8UlIPPj0qPy7rhnNVTo VVJLNB7L4Ft9FkBycHL4/HcYSnLTKV8Ce5/t8DXSq8toS22lCUpSAOSRgV6qrt2PcpUoUoZILYUp SoNQTgZNVW7z1S5BQlX2SDgDvW9u8gx7c4ofeV7I+NVKvX9nUU71H8jCtLsfKV9r6EKJxivZOY81 9AKjgDJrBLmxIRIefTuA+6nia0MzVElsYjJSzuPsqPFRH9qynVjEuoNlifdbjJ3PrDY860E3V7Md xTcZnxCPxE1WbrfHH39jrqnHDzyc1q5E4tuhGz2jxKjXJPEt7LY1jSS5P//S01w1BPnK4ultHZHC tfuUG/FUcdE+dfWHW3W9xGdnyNeFuFw8Rw6DpW7be7ISsYilCxwBCs8zXvLhzk9MCvTccrXknakf 1rK6knakn7PuKgkjIAUoZzwr45hB2q457CsyiMbQnA6GsaUAugnHDuaA+pYSjG4EE8efKiVKO4pw Bj7xr4tZU4TgnoK+pbQpIBVjHMZoD00AGyUoyo81E1jUpLacYBUTxNZ1hTbZSCgg9M1HLOE5PWgM alhQ3BR4dKEAK2knCqyJGUlRAynkK9Ijl7BwdwP9KAwlK9xTnl/WpDSeG1XPr50LSW1EqJODwFSm SFoyBjHMCgIw3LPh7AD1xXwp8FJAAyede0I9txZWeeMUWg89u4e+gIZbBPWlTAwpQyMcaUB3i/3K Ra7YqTGjGQvcE7R0z1NaFWp7qFEBuMR32OfpW11cGxYXHHpvqjbakqUvBIODyOOPGqzZ2rfqB9xu 3XNDq2wFKSA4AAffVsHSovD6k43ty7MpOTUrI2H7UXb+DG/yOf8AjVfdjOIRNlKwtTjm9EVplZyF ABYycc8BQ8x51tbxAh2FttdyujbAcOE53kn5VBgybLcZiIka9trdc+6Nrgye3GvRoKnCLnSi7eUm Zybez/gjLhqjtuxJ8d0xX0JWvPBZH4XU/wA6eSh2waiIjxlXLw9WSnZcFTPhQZiOLaexOOShVwla cTBZMyS/Fwx7SVvKVtQe/P8ApVXXdrM1KcUm9Q20uH7ZhMZa2XO5weR8xW1GvrXyXfql9ObMo1bk xz5DyLY3b5slsyYCg/a5oOUPJTx2k9DivDmr47F8mXplYL0m3JbbR1DvAcfdz59Km/Q2h9RL8GBd 3Y6gN3hcdnAHO0H9ar5sOjkPHdqp0pBxt9TVn55ropaMrxmpX7rK+/w82+pV5uV9TaR1LfsybDDf QmKpXiXS4OKwjJ4lKSedYmHJQanW20TVM6fcX/tLyTuweaUdVZrNDb0FBjlP0s7KcAJQl5Cw1u6E pA/vUtNxt8F5qVLvMGQtxG5gllwIbT/KkJ4H31Vys2oxfzi935+P0XZi3l/uYw0VR1JZiutxYaNh aQNxbR+7w5uK5ntnFSIIchSoswLbdcbSVqbcacSEOHtgHglICR7qis3PTrI2N3SM0gkkhHjcSeZ+ 7xNTUy7EqCZwvCfAS54RXtc+9jPLHnVJ5krOLs/Rllbybn9qrn+5F/yO/wDjXlzVlzQ2pYajKKQS AEO8f+2tF9L6a/x1H+R39KzMLtV0PgW69xlPLylIcW42c+WRXK8LRjvKnZfB/Yvnfkv9slOzbbHk vNFpx1AUpB/CalVEtcV2FbI8Z57xnG0AKc/eNS6+Yq21JZeLnSuBSlKzJFc49MTsqJAtk1ttS4zL yg/t6AgYro9YpMZiYwpiS0h1pYwpCxkGpi7PcrOCnFxfDPz/AANVRIrqZUWf4Do5HByKkzdczruU xW5Um4uq4JaaTjJ88AV1F30Y6Odd8Q2VoE8wlRAP9a29q01ZbIMW22sR/NKePzNW/D9Tzo+zYR2z O3i5UtAaMnwXxe72styVJIYiIV7LSTzKj1P5VUPSHMds2upJnJc9WlNpWy4BkDHAiu4VBulltl7j +Bc4TUpvoFjl8aKSvutjrnhqcqWlbY4lbtci1tlMK7eEhXEp25H5ViVf7rqiYGLezKuclRwCRhCf MnkK6o36MNGtu7/oVpXkVKx+dWKBbINsZDUGI1HQOjacU6FwcsfZ0ErSk2vF9jmZtrfo4sX0vc5a F3mWfDQ5jKGe4SPd1qrzNXx7xNQ7Knh904SkbT8sAV2q86Zs2ofD+loDcvwvubyfZ+RqBG9Hukoc luTHscdDrSgpCvaOCPeaXi+TWtgo1bK7SXZcGp1+hxr0coVKOXGlNLcJGMEc+VUXTTTGpLy1BiPB YzudKQfZT1NdouVsh3iC5CnsJfjufeQrkahWTSlj06t1y029uMt0ALUnJJA99RdW3JrYSFWUZS7F K1pr1uw3UWaFMRFTGaSlQCcniOA+WKog1Db3JAX64lbilhX3Scqz7u9dmmaA0pPluS5dlYefdVuW tRVlR+deGPR1pCM+h9mxRkuNqCkqyrgRy61PR6mdbAxrSzSk/sbiz+MbTGVIUVOKbBUSMc/KqZr/ AF4mwXFq3MTEsOhG9w4ycHkK6ABgYFaC46G0zdpq5twtDMiQ595xZVk/1qqt3OudPNDIm169ziqt R21bpdXMQtROTuSTn38K20n0luy4q4zt3HhLTtICCMjtyrpf1aaM/wAAjfNX60Ho10aCCLBGyPNX 61b8P1OGPs6MU1GbV/U5+nU50lbmUJlJjyJw8f2k5IRyT7s8+PetA9qSBNlLdcmBx15e4+yeJJ91 douGhtMXWUZU6zsPvEAFairkOAHA1hY9HWkIz6HmrFGS42dyVZVwPzp0epM/Z0JJLM7Ip/pBkot9 k00qc4UEtuZ3JxtJSg7cDty+FaXS2rWYMqQ5Em4aQ34j+E/hHLn5muuXnTln1ChpF2gNy0sklsLz 7JPPkfIVAa0BpRiO9HaskdLT+PESCr2sZx186dNtzWeCjKoql3dHJ7rrxq8lHrlyStKCSlISQBn4 V8tmvGrOhaYNxQ14hBWfDJJxy6V1P6tNGf4BG+av1p9WmjP8AjfNX60/D9TL/jlmz53fzc5zBvg1 DdF3J+Z46ISPHfcIwEpTxA4jqcD41Fuuv27wkNyrklTSVbkoCCAP6V1pjQ+mY0GRBZs7CI8kgvNg qwvHEZ41F+rTRn+ARvmr9adHqS/Z8XHLme/Pr8SNohCpeiVqBy0+F+GNmBjv5551yS0XCNIZiQ2n QuStKUBtIJJV2r9DxorEOK3FjtJaZbSEoQnkkdq0du0Fpi1T250K0tNSGzlCwSSD86hZbNM0q4KF SnGDbsiXY7MLLYm4TePF2EuKH4lnmfnXCHbuIFzmQLtuZksvuBSlA4V7ROa/Rtaa8aRsF/c8W52x l9wDG8jCvmKJriRpXw0K1NQfCOQR/SC7b4YYjXlSGk8kJTn8xXq0Qb9rqd/orTrUVR+1nSM8uyR1 NdPh+jjSUF0OtWZkrByFLJVj+tWRplphsNstpbQOSUjAFTeC4MYYCnG2ZuVvL2OTah1hH0y6NNQL j4bMNrwnCrKlKJ58fjVUY1Nbo0hEhuYjxG1BSSpJOCOR5V2OT6PdJTJLkmRZGHHXFblrUVZJ+dY/ q00Z/gEb5q/WnR3uRVwEak87k/t8DmU30huXZj1N+6eIhZHsJQfaPTpUvUV3ZsFojWN93D4R6y+2 BxyeITw7eddGj+jzSMSS3IYsUdDrSgtCsqOCOIPOpbukdPvXNy5u2tlyW4CFurySQRg8M45U6PUe 4Rd80m7nHYmvRDta4jFxCWHActFGTx5jlVl9HNjlXWei8yYy40GOcx0rGFOq6Kx0Aq4RPR1pOFK9 ZZszHiZyN2SB8KsiEJQgIQkJSBgADAFG4rgtSwMKclJtu3F+x6pSlUO4UpSgNRqM4gtnp4nH5Gqf KusOI0pbjyfZ6A8atWtIb83S8xMZRDzaPETjrjiR8q4SuW+vBeUlSeozxFevg6+WllRjOneVz//T 9q1lHCSpEdR44BJqBN1dJltFtgeCD25/Oqw7MbQrCPbQfmK+Imtp9otnPSup1ZvllFCKNkmS4+6f EWdx5En86wSJjEUqC3g4/wA89B5VGTNKyQQDw5pFQZTrIBaDa9+ck1hJmiIzriluqcCsEnPOvKlr XxWonzNeTUqI+0he91oO4GAknAHnVATrXu9XVvB2k8AK2TbKTnKQM8hmtTDU5uWG2gr2vZyeVbFU 5iO3sdbT43UDiB76upCxlcZU2MZCjnlmvqWiniXMDseVQm5MV7ctx7aR0OeJ8q+bl8ClSVg8dijy qxBKdShQ2+IgAedY0NjbuUtKgDw486hkK8XCilIJ59BUl2NFKfspQWrhnhgClwZVICgCkcTyAVyr 4W22QPEWgKUeR61jDbaWnFeM2lWfZSATUdC1BYU4yhxIPU8akElxTY4LUlR8ulfUICm8jxXAD0Sa xTVIeKVNtpScZyk9O1SU3C5swk+EsNsjkTjNNu4I5fbZBJbKSe5qRGdZcIXvKcc68m3OP2oznZDW wKAA3Ddn3Vi8OHHZKjOKlpIIShPA1VuxNiS85GkYwFIUe/KoaJDkdRTuQeOOBrXSJjshZPHAPDA6 VITBJgiYFgDJAGck1CYJhc8VZ3gJ3cdyTwNYzJYCCElRKeABPA1DKw23gpUc9DXktMEpw7gnnkcB ViCZ9KqTwDYwPOlRD4TZKQtsgdcGlAd49I3+5cz/AKfzqmeh/wD9Xn/+yn8zVz9I3+5cz/p/OqL6 LZKYcm6yVnCWowUf616GBTl7KqpefsYVP+1GD0n3Q3DVHqaCSiKkIA/mPE/nWgfhytM3qKXSA4gN vDHQEZqOqY9MvS5wZU+4t0u7ME545xU7U95mX6U3MlwPVlNoDWQkgHHv99fRUaToxhRS6bb/AB/t zlbUry7l+9Jdwcl6LgPsqIZkuJUvz9k4H51RNOxNMvsuqvtwejuAgIS22SCO+RVjgaliI9HCGLhA M5DUnwVJKsbcjKTn4H5VrLRE0Xc0uqmSJFtUk+yhSioEe8Vw4ZPD0JUmmkm947vn+9jSXVJM8QdK syrun6KvsFSA4C2FrUhwj3Ec/LNbLWWgYenbIq4My3XXC4kbVAY41TrgmLHuribU+46whf2ThG1R 866r6RitWgWS5ncVN5zz6VOJqVqdehae0nZq2/zEYxcZbcFL0No+NqoSvWJK2fBxjYnOc1j19aW7 JcYlvacU4hmOAFKGCasvod5XH/prU+ln/edv/wBgVSGIqy9qOk30pcfIlxSo37mO1WPRb9rju3C+ LYkrQC42Ek7T8BXvUsC0W/R7X0JMVMjPS1FTikkYUABjjXi12/QjtrjuT50hEpSB4qUkgBXXpWTV AsrWj2I9ikLejolKK9/MEgVe8tePVJ9XDW30/wAEf+XwedEaGiaotj8p+U4ypp4tgJTnIwD/AHrS aqsR0vfjDakFzCEuIWOBGc/pVn9Her7Pp60SI1xfU244/vSA2pWRtA6Dyqva4vsXUGolzIYV4KW0 tpKhjdjJzj41rRlinjpxnfTtt47ESUNNNcnadNyXJenYMh05WtkZPetnWo0qhTel7chYKVBkZBrb 18HiElWnby/qehHhClKViSKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKU pQClKUApSlAKUpQClKUApSlAKUpQH//U7DSlK8wuKUpQHwgKSUkZBGCK4lrnRi7NdXJLLobhvqKm +BO09jXbqhXa0xb1b3IUxG5twY8we4rSnUcGD85KhqJ2trJcSPaBHP3V4RFlDiI7yxn9w10+5aSj WibvdjvONEYQ43gfOsPqNsVgJYmDHQOCvShmnG6RRuK5Zzlx11hOFx1ozwBUnFQnypxw8ck81HrX QL43aIUbx1Q5C1A/jWDVBuEht54lhjwEHiE5yfjVWmnuibprZnlploMuOuLwU42J/eNG/BUMuApI HAjvU/TsSJJlFU9JLCeeFYNXafbdPJtWY1td3uEJSrcOZNSoyfCF15KS0+ttjw4hC3XCMkDj7hWR 2wXdthch+E8lscVLIrZXizMw7tFbixnm2iBuVz41a41uiOwQzO9aeJHE7uYqYwleyRGaPdnMShO8 Ecu1ZlKOwDCgR5V06NZtPRjxtbqh1yc1KciWBTgX9HOqOMKq2SouxGeD7nJkl1R/1KlAc+Br0Mlv CUq58sV1dDNjG9KbW4E9MKFeVRdPYx9FK4c/aFTkqeBmh5OUbXCkkIXnnyryFOFB5kjr2rrSYunE IJTbFhZ4EbuGKjSW9OxEDFqxvUARuBo4zSu0M0Xtc5m0U81JUrPwrG84or2pKggchmuxtWuwyMbb chHDmpYFZUaWsLqiEwmFHPH7UVDjNcolOL4ZxdDZKCVLIHYUdZAZKkjOOPPjXYpWlbQp9CIzcNCS PtMuZ5dBVc1TaotijlxhiItb48NAbWSoE+WKo83gtt5OdALzvAI2962jaTIipQhOxZOdoNby0aVu rSjKfjMu70/cdVW1TbZkWUC5Cith8+zjjg9qs6dR9iuaPkpTsRzwg4lClJPJWOBqIuHITkmO5w5k JOK6cLdNU3hTDGwdA3wFZmbPMcBbbcUEq5jw8D5mradS26IU4Pucmwr+GflSuv8A7Hg8TGQT33p/ WlVyz/ti2xeb7ZW77anbe+tTaHMZUnnwquQvRnBgRZkdma/tloCFnAyBSlRSq1aMMkJNIq4pu7RJ 076Pbdp24Ga04t9zbtHiAYFbW/6bh6htphSU7E7gpK0AZSaUpKrVnUVWU3mXcKKStY0lu9Glsgx5 UZb7z7MlICkK4AEcj7xk1APoftWeE2QPgKUrdYzEptqo9/h9iunDwTLZ6LLLb5bclxTsktq3BKzg Z8xW+1Dp1jUVrNvkLU0jcFAo5jFKVjOtWnUVSU22uCVGKVkiLpjR0XSwf9Wecd8bGd/TFRNR+j+H qS5euyZLza9oTtSBjhSlFWqqq6ym8z7jLG1rbGq+p+1f8dI+QqSPRbbk21UETH9infEJwOeMUpWz xuKfNR/t9iNOHgjfU/av+OkfIVJt/opssKWiQ647J2HIQvgM+dKUeNxUlZ1GNOHguqW9iQlKQEgY AHSvu09qUrztCJpcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPal KaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptP alKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2p tPalKaERc//V7FtPam09qUrDQiTcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalK aERcbT2ptPalKaERc8PRkSGi26gKSrmDVUvGkZWxa7WQVHkgkDHxNKVtSvS/KykoqXJSJmhdaTEL S9Bbc3//AHDeB/WtO56KtYqWT9GNn/8AZb/8qUq7k3yW4PTXov1m2kpFuQASD/tLfH/uq0W7RepU tRW5cIAIWVuHx0HjyHWlKmM3Ehq6Nq/pS7yXEgwwhKDkKLqSfzqR+zV2AATG5f8ANT+tKVqsRNGT oxex6Tpy7AnMYcurqf1r4rTt4x7MVP8A8if1pSp95mFRijCxpi8pkvlyMnYo+yfFSf71I/Zi4/8A DJ/zp/WlKhYiaJdGLdz6nTE8/ejpHuWn9ajr0ZMfeUt9nhgbQFp4Y+NKVEq8pKzJjTjHdE1Om5qA PsVKwP30/rUliwLScuwnCfJ1P60pUvETZCpJcHp3TxUQW4q0jqPET+tQpekVS32FuRAtLSt2FLTx 7daUp7xK1idNEr6DljlG/wC9P61il6dlvxiExsOIwpv208/nSlS8TNjTRNTaZewZbUkkDgFJ59et fE2icF5ysDPLck/3pSo94mTkRnECeBgKX/2/rSlKrrS8InKf/9k= ------=_NextPart_000_0007_01C0C5B3.62DC5A80 Content-Type: application/octet-stream; name="oledata.mso" Content-Transfer-Encoding: base64 Content-ID: ADoAAHic7bt1UFxB9O+Ju7tLgkNwdxicYAGCuxPcJejgTnCXEAiB4MHdIcjgJLgPboMNDCz5vfdq f/X+2Levdrdqd+t9b33unZo70326z+nTfWt6QNP4myX1FFtw/50k4RDhnl/Q4VD+03vw/5X/EB7c 6/2Xl38v/9v1n17+l/4/Jdgr//yH+Oo7pFeQX/nnc9RX0F5BfwXjFcxXsF7BfgXnFdz/EgJw+K8Q vEL4CtErxK+QvEL6Ctkr5K9QvEL5X+Pkf+n/XfoA5/p6eMHRwcnDubxePeD8//tU8H8okteI+W9l wf8PPitNnCCRO9QL/5/zhRkcDxw3HD+c8OvBDycEJ/J65v2fqJ8MDgH+P7fn/8x3EF5Z/p+p5H+g /9n6/+/W/6X6ORHg/PJPCzA7nfW8FzUIilNmM/XiE/l12NDRWNSKw/Ep6ItLJJz3PjjjRtz+cB5H PmGMvI9RWSCgjTH8qaIS81MQrduSvTImVBddIYLwPbsASxYeslHit3mFgR1T6Cns9ObUNZFnafVT z1TbnW2h2DDsZ+Otsefkr0XJ657hbKuJh1QxGv87jWcIctEqutT0mojrwXwWTfM13oovDST/LEvx 5CRf7BE2n9UN+eZyByxJs+qeD56KKUoDoU4WWlQw20E1T0LIIJ/BWQBaHBgGWE+2J84srEBRlhY5 iB/CA6qNCsaG1IM2kGCGYGTQBlYQFeS4mFSWFhcWAHYNk4SHYoM2MILeQN6Ah0DMMkVYMGdQhXQR RhAFpAUQggo5BLlGmCJ0EUGxwHwlr4UUgOxkirBhJqDfxT2U0PxD30C7g7ySLEAILsQfdACoDe2G g8qCzyOfJcEZIFKZHmToT5AJuEWmhxpKDdrADeKBoEEYQecRz/Bg2ohgUkhHSSAoS9YMpQsD6gG+ k+4hDSCCRheHkEEVizdQYcHgBRBtaDdCABZUBVxVXBv1rAXGlU0GmCF2oUABoCVZM9QuEuhIcQgV lLk4hCCAGqIh88IPmZY2w4VtOayE4L2YxttdIsw/Gs4LiFg+TS9tcnU/cqpBgdcVXYNTU9OXO6QB f2pDTpPFD0ZcH9s8mRg5i7wZDjbVux0DFqpg+T2eZyLPQTSBq1Y+7ddStEsvjqu/XUc8hZshD7SR 1xqcNldC/LdPaVcdjafXtgddBL4vjxUa3ofX5xyBsIPJv2IPIgKBLs93G8E0wlIHeDtdRWaw/WhJ v5O7c6rPnlNiRRuw+5YXm/tzsULfyeNgMKrPeF7P80mR/+QyIfZzfsviTFrPUHTtWhTfXcVxSA8+ P0wQBP5FN/VLFTJCWvySJx42Rmp7RU3mObFLeaFt5usi+pIh/HhXwlXYtLPwdsVtCwO2AbQ9XZuf bjjYgEyLHrw0gqRQYdcg2rBu+AAJCJf0yxsIl8wLLyQR1IMQRAiplX2hgNQCQoigQsUhhFArUA8y bAXUgxjEB0EBJwM2ELrwoR/BaqAeTBg6WEP2hQZyDgjBCOCDOINC4ILEINSgEDSYIigEC8YMCsHo QoZOF0thw8jBUtHdhAHsELuSFwQoJ7gnopsSWlt8HtZNCpUE96xbSN24Fr+QQM+Lz6PXUWBwkA2A a/QzKhi15AUfSg+OLnkEiRXThj0zgwtARdHPCBApgBQGTB4cEraODCOD9Mi+UEJ2QIQgDenzKHWs 2sDiF2qoVHFReDAFVKqkBwlmA34Jf3YFd4FeEGHEYLviHgwYB2Sp+IUUegx6LH7Bhb6U3AFe+0Uc ggniktWQ1ZCpMCPyCk7LyvIsOZChRQrihtCDuICSRFBu8AOIK1qSCkoMqo0IJobgg/73cSYZfU5w ejBx/4AOGZhe6XJYpNTgqIhduDE5uMK+ya/ERlVFKuqhtd2Q6ygacPF/dJA6rYA1Fd16Zks9VqgL OMyWKad57yY9nj+CT8+7jqldW58zuDYCZ1rOJA+u3k6H0D/3nDhoRBtfqk33v5BKbZUym963ciR3 mV9JSQUdadsOJXWdsOlEIl6+PKuN0Es9WkklOI9i14ZQ3216rX5HG4+wvCfWLSF4SnZ2Mr6Z1TMJ XLvhEFGQ9PyNTeQUcl/eouSG2srzAFnMMyj+iDj+Pn7r5dG1w+b2Ex7xtYRBO4WfyfJUP6/8paK/ pTL6RoQNLq32abZlnuj5B9KVV8uZ0jvOnsmg+ttRFwhTnCF9GZt0zegyHimjhI/G+nBXCfGFEc5X OEqMqJF89O/F4svmCuHoFOmRJt6gXr54UJ18j6DQxxXxe+yQiniJdkMUaXvZ0i56VCI5me7vMNzK wmTre2nFSz8QnNat9l9PCDyLvJ+Tc03+9EfKZwJ571o9v3RUhgiH0Z1DxONomHJAUaLtUFhl+2Ki 1dhOOLD+DzGr6UwKkdHCPzsfbVHau1D88EO40ZO77xmfdLMDnyI3UpAfHdC/APkRDehxpkcBrtMI O1sZkU/bBh9r4hgiOw+HW7RldV1qb8j4kUFy5d89NB2BGTs4gWFxoE5qqBcos3MPJaQ8reb5SPWu 6Akti1wAKgYfU6olaKHIJoJosCxFt13+Y9LOLse0UXFpeiCLrei/GMOPuIb3tV5BKSw0xYqrA+MF MbyXSpkUjcUK7+SrLxEE6dGN6NNALf+iHKDAJgHYlMQdbquGs728U9u6//e4poJITbUuqmLOv7GI Ly/f0us2azxZ1M54jFUUrN11Y0uQ03szYoiEgWnHP6ryBYuY3dwLyWu+TItqkNCs/Nkmjz81L0mI SNBhNuJqTOyWHUL/w54GwZphja/oxjj0vJ6UKL+0Dh6+e+3AOMnZ6UK6nfOQwhunYX+T+ufb2n4Z WvZqY63IxjJiNnhIbcBhVVQKG876Aq3u/FTptzGLxdgWz970tLTMyV4tF0McdBpaa8uczKdSeRwc 8oBGoxkRwXgbphFO1fwqLmJ3BQ831pqfozJrcUxqueKhkd2g9kt5ovF69Ik+ECJqUAwIeZguURHy sOzGjR9/NlFWXxmas3AifGDOpBd5DgDVuqU0Edz0cqxKq8BZkvnLFbTytQYiGPNEKODzsuwLWFdx 96byR9xf/3RzE3Bzc/up1P5AZpwh/MkdL2NHTiaJugO2Rms4Hkwl6prGpEZjf1U4HZwL47uUwb9B oB7l0OcHj4VJtSusUbPBP5PR/YjR1P1NgUb6lGxJ5VxWAJ4cddbkG1biNKsdnJmWsKeMUHuXgL7u WSvIktaaejhs4huSIG5JTixtlojhXs+PjsgStItPa3Pu4tOyvpAYh9BHGP98lvzOBjfTXbA/NDdi hEG8TePXtteaGRC4zEfEqtP5vmnuN1bYL/c65TTU6s5rbdqQuUTXJbRilEM2jyfipVDLXtq1EgqS 3WYnQOgX47mw/sQGZnufMDSsirYnZL3SK039WxOSP8YPX0Nv4nTYLLRnGLgUtvFE0y+qUoTQhY6+ e51LmIVu6rnhEc8mlY+Je63UF6HVtIgv9xOEjBPjhLOKRNiGvV/LE/nLh8uBI2q1tElnpYKaRocy BeYlovYChySo5fKL+MEjbUBr7kwe94UQIsFFI2jjXy9+LZkkTz39yOrG2LW0/66ALg9Ag4v+8iNf BI1o/84Cscgyz2ev5fOB6gyZFJIsKMIWTZQunoPrdCRqLq5iUbGNzcDRgIyLbZ3402bsZ9wUxfVL Q5Z2h9b3q0JRZDxfkaQHu0o3qYEW8n3FmQbU6+wIz5l75iIke0eqBzfWKj5hjC5QtQZxl073YB13 HXfA7SYIzaGg80ej+XdBHEGHq9B08pnyVXqM+ui1++nzFw98jNVpxIWWRlVcvLLImGQkASrVPDRK PNmw4pKgwatBJEldq6cTAt097yO9F3u1y0wuRmEVoXEzeq8NvzH+iPpJUkR0YTS0K87Ll7hN/bi0 pSLvAqA8J2jESBz0rsFGm1iQT1KfqKH6nxeZdpCCeN2u12r9GdvVXrNxTLs2hi+JlGVWFS+3nTVB ezLRZzlcjXZ/CgQkxa23NEnYFIoc07CjF1iZ+Ed+9+rdAoN8ud+ZI4LfTJPHiVb2Tw8egZxUekcZ O3pSWsK1U9st/DFSZ0NUHDakmanuN99E3Q7JKFXVV0lQ05xXhJdfsw6H6+MgKhwFlB5exmtUNjki qPKF7ahSHLUN/MfqHkzIdfIHCB/Ty1jPUkuHe4ByW0xDASRq0Oh0U/mAhiD43t7nt5IeT4SfDs5v TNnRUgbfWb6xT1LQ7ay6fs1kjEBnaAYlC2JjUs3Ua944S8c9h2vnFEwmsfTzd8s+eisk4rDWWPA8 SIqTT+zG5x1MEm4hWgGehRdcgyvNnf8P6wCpvsFHO17jlTuMlva/KqNxJeoQI63CBzNR2hjCc7eV iFXY7L9kc1T2OyEGXG1GDfmLVBmReET7fXMrxqZEIq/v/+Mq/JYKddOPTqi1iJVmGWglYkMV8WTJ UYW75e5I7XLBBMZT66YY/25BYGmyRx3D2XnQ+UYCOYmGoIE338ZG+6mEC8uAXA7QH5dfGi+yvKd5 j2VLuUlXha1MKNr8IJ/H7/AhAzf48houMe9dP+pB4Bb/37IE3xP1ogmulwycD/5yIl7geW14tc79 NUk1HdT6aZOdpMGEzVArF9DHjHt57FP0yHgxgTnhXC0OOlFUq0xfmp4oX8TF8ur3EtTdOWDpqwIv kq9/oi68ntqSt83HBCiSeSvs+3Tygw2zbkS+54Ay40h1IxOBr4myd+/neQAtCyA1WttF+u9HHOo2 Tlb4cF4L/hJ7OuqCms3iLHwie1AleViFpTv+eLdZWrBYo+xvag22opmPuybDRe/F+6r83gjRADOY 44WFMnx7zxNctWVXnqZqh/tKvs+0jryR50NFLOF1kQ3jSfht47ZOLLz8aQEEEf434leQ8coYlVJh 8iF0PERPgX9mmCPGei/L0GdS5r8afcD5MKMdi/jXViIMrQT+6KUg4SywOs+7T0uEAXBbfBbaoU// YPwXD7UgyoPpWIWvf1CG/MRjd/btjyHJeSWDEXa3FGQmVYEFWH6BtlfIKE1+OMFjDkCrl+X23st5 CCEzZVQuYaK6RCcVfYd3ckCFcMh6jHUDF8bV9Dq4jDJWzjrO79lrSrimnIXalz6HlYGqop9pS6VJ H6QAdd8jYG3tQfA6KHS/vw9Ms5Thj8e+BwpQnDRgrDUgC+nQmQXWoQr0KwnxupqkAuUxHo/lUoH3 C4i2kGbpy5lhUN0J+W2FVBRQfsneJ3Axioh0vAH9UXs3kaunf34j7Pxfx8TvqQl1S9xl/vEEu4ha 9Gt3jsmQe53JeCMEvLFAc3FXGRfLbfy9ABf5BcXUOtANTTQJ185S2qJVCj4JPpeGL+3c4ruKHoOq oLAs11km3770z361B5di7UDjRMo6Gv8q2/s5WhG3ZnMPdaMsi+Z6jqXMgifPdp+tRZHszFzH2bKx P8znT/MjInw+zSXzX6G18EiG/ilUDBh7s+lh+0jzgqiojE0KiKgZb4vT4W8P5BRmPMFMRj7eUfFX 7Ff9YYV/CZ04F9JGsNcsKN/zYXuhMBFAaOgH+lvpl83fVwYmIM7x/bp2NuijWaXdotjrzc8yG2cw tuGJ2Tvt+6a1Mr1v/88ANm4UxL9UJCWB0/sK3L+IXqZ7h+IrjtVY7vm4RvOySt3uysW/eaHwhirK lPRhec2xJvwZUyCsZV/MX4kdUlldxRzyInken0CGcsPWeZnxxRA/AHTs3AbiPk3wzF+NrH85QZjE G4RdcvUq4ajvU1DnUjFdlpL6NIyWjMNlX6VYyjfb4aqjvpHY3Qe9KL0O1gYJ/2k7qUR14jWOuTcx seL8+7aq4PY1+ZxD7aZHqhjycEBOaFmJUSTRTMpAeGHfCIcfek2ql0/m5ZeMT70LEQNawTEo8NSV +adpFmxTMhr1dsq5LRMhg5RK9qGzykgQKgZEiyk0vZ+zOyztFvFMzs2oYxxmb61Ydmf1RvT17+V6 PEjfTcB3fozT5eWIO2WNo+Ut/yyy+C9Q7So7b0zOrXDH1tcTMT9IO9i3Ab2ZLWhsrGrVA7mrJVLV ZM8PuM4I2MFq3s1fSO1/2GF8j4zAbxsoyAlywWr4w7LZ2FSCUoq3sgqqKDUi4ezdFZ+F0fdRfzV2 TfXsqYvMpY9O1CtL0fzeMfj49TAIDOKMGrzwaWsbN1BG/xQEL4cawBTkclFKVJFJzVbtXU1UP0aY KGMnv52/NRDPpMdZxtm96OaZKOjg4YPZbrZvEMj1Q1KbJPp7SbAAmdwZ2Xw6R2GklntiVCe6Bj0D KlAgDrq2Mfox1WCbokUCJoV1IFLBL1TJBNBu1KV3jcRFPio/5NgKkh729vPcfDdLJSCOnAVhpCJ+ I2aN6/EZZNeo9evsxGbS9P0KFGHCSkZkVXAi3cZGIY0Tt5+04uJ9b07HF0Yg6i+pgwSOkbZAn6Ou ed7xZVveSMHFVAPT4Sbsf+kw/2c/GrkN+f19FgcrjE3LNHTQJFWl3deufN+uyi2FZR7JRTwiqoT5 aickXPlZQfAD9seq1g38Ykg7nlEp5gGyR0VHzC9qHPqpGBPnUbZ0fIqpb6SxeAOAlY9CaloKVwxm TmmJZrdNpmAnP9zLOeIR/xvqbNSmVa+nXS/Yqvfu84qLRYC5m/nU9VyFXc6pgYX3/ZRont7ykOfs 0g8UJxWvOMCL9oLx2b+pd/36j0at9orz3Vq0ig4928smqVTPpeAOx3eDBtu8vNb7RJOjyy3WCMUQ ltMRX4DZvvPnu0wCfqAxoQA/T1YK85EIalAICe9v7c4MeX2sL+5cLAXBTbGtIxVTt+jqJqOWd3ui Q14m/MdUnz2Qo/1lDrvStZTtKuMcyFKPdz22pDgT2FfSDrwbRTvftebsWyg13H+WBaZQ++d8G36s b9d3T6EHRJRnzfiau1jcNM7rEU0djJ0m0WGExqKlzQjx+gw1D4ST8i3ZjIwnmqhn1YXoU+2ekByE nAJ9Z6etuPyomNMUs3i7r9tuJ10HhZiXfM7aDjzimPPdwW6fca/DRMXvvt+z42Dflm+3RznO4ebq tOWrBB5JtRb1wtd4WYzV8apy9mJuOKpeaF792uns8ul+SLxRTG7cNAShE3HBn45UAN4bXx89GVz7 bKkemzk47eQZP2w7G7mE3EUvB3rK693e4silTm0RREhKUFddXUUGaNBUSzGqbGrnQyrnDhXGJOX7 jRdtaO1vUF3zpshQzuWOzJej6DRwM9DFCVRWG+a2tgvMmQEfRjy6/N7ZJb+Eqa41GRfqPpjoampm /1ZvI3b8SVXiz50FX4r0BreTYw+Xk/grL0OF5B3+eNfd5zAGjI9GDJvzglSfiNVJJU6QPzvuuhge OA1nd0e24otUC5N7hVoFUD6NFRMa2TuXvLAScL9s4uLBHIfZTdfYFbfiKcZRcc4gHZ/mSUSvtzX0 2edssEajgtFaIzk+aqizZlrQdndlEm3wyOZcJKXSSmAkOGDlpvo3PS88NA6fUgScCT+WLFfQdsG6 fLt2sMGlDlOaVOR/ZI4qUadyzcVFMTKk9NEz2lql8Fs6lOMLl+mQ6vWr+//WkohOhWNdTsWe12zi pX7pqqYwgthsL/SzQ1mjS+DDO4wiV90aZyaVLfixLEnkN8LU/gJzP9hCq9CGCrqaYnAstv9+Hlp3 iUSb/gHSEvnak+XpPnep7KGo5+Fjbhpsyap8HUibUPyC24UEtc4C5WwmV4UnwCAfkwf0vrtVnlFQ 4drJFCQRe8Viin3wQiDu+SNajmFoxCwWskpFyQYMaP775SIpLLA1rRu3zytj8wyZj1sK+rn9gmWp s3278dYW3BF9ChTCkvI8nixq20NBe4sNJ8c7ow0nvstrR4WE0P4OwFwmXlrIqfW+GIRUBqXtrxep yAMxpfryh/AzqCu7CxKNBvbXFjpu3TH7iQFYtIaesHzz9gQmcq3M1spu9F+fKwyrXZgDQlM8NW3T eSkU4ijVrjTignD62sjtP32wPzDHPpxr8NQdtvUl7hM3HiETf/XxHxcweeFFF27bbtR1XHgyDezE E2bTYU8KzqBh1Ar+LnsS7TdCZjrTc1DAReGIzuB3ACLLZ8+sP2s3zSub5UmqLmGBowI31pG5xLiw npk+9Nd2BFe9k6F8gsFrICWXm9qCZ5Z6lAIg4ERYc62/dwlKPmOLnJl403WolhLa7e5i3uEg0ILM if2mJg53Rt03jYDWkAJ95Wj48AErjPZmlBxNzMegWJ5CFEni3FWAV5Rq7d2EI6yMPKnP8xhQddZ2 sxSLZ1n2e0YV2dUcrqWC80iPoqKBkZiozLnt/VtUB37WTs/O9xXab93VfWKPgHhNyKxU4Wlma1dH t78jxGeKCpiwcFCMSwnUgwbcCh4Uq8jbTBpxr3AJqbk6vmLAA2Oi7tPIKX4ApPi6Hz9ErjOW7o/T g4+fyu5nSTrEvJdu/jJIQeNdydu1rKQ/Hj7MhPJIPD323H8Spz47BUW6djBOCVMkt72fYSjzDXr6 ekdKr9v6L/hrepZmZMWVUB3CgAUgca/oQqqKkhVfVc1svI/WlcQiYFUtaw1WRIca8CwcIKojWRUl QqQyOuUgndQOETJQW/Ie/7OF8ezKJn118Jx4PlPaGn6t+TzeFdvbJuG6450KXp0CYKFUa/rEO2G+ QSJbcnr7QwLRTHoW+hQm1+tekfsk6rx7qT3q06fvsvTC2bETiHoxTBL+mw/Az1u1nV246Dj0bbJC srXzfj+iMODlHpRGfJc2zS8TzhyogPTpla0KS/y6w3X6dDTbdNYMH/RcYnusNn+V2ed8Ed56dhN3 +XMbExWgnyeAfhzuSt0tXoGI2vj9bs7vTQlh9hvpezhyNWafIckcnyBQCpVOeKHjxzua6jrxKDIF boBmleRNn2a00/fJyQ+1PrW0BruB/X/ivUmMaBAs1wgpCVCoy9umujiXBTqb7pfMf1n7aCUr+/DZ kmMDSmmFwGqn2Kdh9rZYUXiaFafoRTv8OkpE6nkzb4vblJVXMtLucL8Awbd6AySB3JBA+Egw1jGJ zLdkyywWooaBiggKOpeOLa4uP/R1E2De0F/ZX3WKbOdYGRQQrXoso3BBruowOk4PTdLuvkyi5AwD DBwciUkq4zgJ9k7NWMUGjKUeUfHj8bCH+ZvLs4GtzZ9CqO53hzNEvGxEHuUU5AKBc15DcB1xyn6X XqfMN1tJyalUWEmr5uMPxRUrXmw7gKIkQaFf/Z96Zz6o26DY4Dec3WMjTahhZ70N5JPjJ1ab9+4W PFvm8JMDZMctvhNdLMprwgxgtdZND55sLhMzXNPRKVan+xUag+BLw7H0Ee8pB8u9D17IOoqbxGQI lEGcSoQHLMQwcb/slLMga73UfhSmQA7pB7Bw7SPsauMm+hMjZ48yj6Mj9w0dCPH5ztuH3g3L49BI q/bXT4ixrC87EZRegLR7sbNKEgXT+nMU7y+5xkT1SI+WjAhzU4ue+YpuVoEPRtr5Ex68e3SiQrOa zRZorDN0pGLWZzJ6E8aCzGsgDLUJzIv6UVAVKvqi5ozIYTbnX1+c3O0pzd76VjsZjpyfFtKgWvYV 9z8P+cc6PDkrBG3K1/XijgW1avwoiO6y1vFfxzXJvtpw/KEnKX5pZOc+KI5cRSok6fw616r9PO7v gLOv4WQDzxoXpJeK2fRDVvkct6/W0k6/qTGugKBj0Nx2seMTUQ3SrjUuzZTYl2zAl7xEwYNe+h49 i/d8Gwr8unVkY32JLQxU8Y/G9WeebocgdtBo5sEXdjOT+36wWKhY/unit/XwxdTk9ef1jFOZCvqI ZguyrSsGXxFwkhfmo2S3UVyWyJVh1xLAnheVGomG7dCL0oovil4qWill03ZCOZLeWif72tM/Et7O tA0FQ+B7Lg+Fk0ImrrtcoOCI+gWB6yCN17HrMOm1tG27fNXP3+4gmr8J8jiO0xFeMGtA4upNjP/N N0MKINJEpyRMRD/hPdlRR1QAs/dbsF1cwVYDtrQG8fos+5r32q+8a+z4LOn4iW9fWITkZjUBrfzn Tr+Hp7yXPK8IU4J2JYx7GDN8sZoqaB7I13mGexAO5wFRK1C9A4FGQzW5lBEe0incOvyHXerHKnIg fCTR3ygXGscoR5/+wDlykySfMV9elIF8vlxe+cGxnSBAvqghRuGYQum1PGVfCHgC42s04GIcMWTd idm8jnCthsrRx6WkaH6LbB3eY2CxHtVukH4UjqlfvlpZGW+qsrWBgI+kw72L0LTNXRMxhsaVULij KMbAmp+yMySK46fRyd5RdEThSJuBV9HiJ3YFl4iwvGUTX/woLH/qo6r4tRuQLpeAr8qMDnrOJ3D9 jFPs8NrwPPczTc3G6+xLr8JJdYV5TtZqperuFqLSfhGJjvtpr+oa30KHYD8T3XvR7EmQ0XcNVq/7 VMpIyyoPiC4tsbtt8/RYKl+0zsxnXvMMyHm7VmY3hC9xEH5LpQZfGh5uxW7FeC8iMX8QwwiHIsYp bOI9bN5ia6U5nHvrx2Yp6Nr4Ll9E6aARcX7StGfd+0TGN+5Mw663cLDG66trIxGQXsn7kpt+T0as oAQrNOdtLYD53Qj9/FRnOS96RxSrUznp74AA1VRjzJ2xBZkH8PRzaFALN1GJH5HpxzUWzK8RJzXx Ii4ycTxh8mmgwCXoKqu8NmNKkzAgvsye4EfUE75oDf3dVe478oARFibbKDGSqCDPnEgcuxJuLGXX o6PbCqssUSRK+Q6BTkonwFnrS7e3iD4vgYDXKEtufttmnRvhYYMLo4za7HZQsA83jgjl9Wbu3E+0 PjjWoXa/a4Sf+cCpApO8eKYlrsKousyUL+dEv1228lE8Bhd/XnwUSYd90XuemtCfywkjRFNvRAvM CQiSvWSJDPCQEv4FQYDYMqg7qqd1TSSMBJuzoQvHk0pcm3yV97XEK2bArSYSq6Fa/fJN2LoZxEA1 7u45EXnW7qHg9cgJ8KwAX3kCAdg3K/r8PtlLmbUxQX71ysctZfdJWrvmwnp+IyySg+3KrRtDas6H koipkokTKZhBFm+F3qhi4XNqKKj/LS8Z+MqKxZvtEbiGcEyjTABH4dTylhVvSLEPAZVd/XdpPK++ RUll0qGb3214e4iu9kirf8cnz/77ChHReYsDLuL+Tz7TKNk0hwdqCVn+1vFi8g7W3s8Kk7b73SNU Z7OUnNbPwalBc+Pxi0XvpQrAU16lSr2VLz6SpysOXJjkTemvTePxkcEvBgjMbYmk75A4sudGYDY2 H581WAI9sTPyB3kElHYXAUW0R6+jsk94b4CNyXYbX2edJ64FUqnhD71X/isUXCTgab7WLJGtmYPd qjCSE1dGzzARlvl4tNmnTa5YMjc5s5MWTvJz1MkERUdE4Q46IxyYSSzM7zI67q1QJMo0AdtOuNZ+ FN8KFunjv5Xp/WoCSWXS9bxzdUM7//bUd/nmbt+J/6J8pwlVHpDj4OLW9lAOu97qf940T0q7rPtg ECK/ZL8uJWTS8WL3Nkphes4gbm0pW4SsrRtOfpUj4GmYwZEYNZtti9G4pdRe4pTZnqTWxmLKO6i+ sOQk+ZL6JJviiU128SHTsN83wU7yuE0e5+xOUY8dX+3Hu0YxTFL5bxxUjFQh3vkXP3H6gOS0qxWf vH6dxNpoKjHr8iy8GewlMkmaYmoJDZ8h0icE5ThXiXswa8//PvFJ8yFR6+QesCODm2z8XvnnKpKE cjMaiZAXqTfm0GL5MibRKiqzdlpmu5Y0kpc6lnKg3B6U47CfNKkJyN7maJWOdnRDiFsjfHZUuZgz fW2RLJGHn/pbLl7sOOosDVJtonx2WuJtXVnfREk+T801w28d5Rz86Hu12qTUAYeRMD7KgYyfTeAK HLNxhjJuTQapLAGjLxs8cbE5nwczfBMs5wmv3x+zQFgbGos896qArO1uLqifR8gCAuBe5yACgfbO +6aJUgucQvSulhUbRah2yt5lhzk9yhxXoGTj4e6Obx92hx1hhOgnpErzSuTb09uKIZ/Qdt3sip3F +pJJfxZNtXdnuL4Ov2rHnbK52gDEbHyUSCdGRgCqiUbfj9WdcJNiqgqW62ZlzYaBCm51ZLu9Wz8w OvuhyGfBHOEZb42JSRf7m+Lhkrbtc4XB2UavTbhL7rfsFs14+0vp8zCLCsgWMtzNlLoxW0ppMEKS F43gPzk42RzhKgOgWq3ttsLqQt461qpXHnjz1qLXpaCKUuXLHvCWORAOnQjVgmDdAzAqNJ/A34Td 6L0vjFnQBiJ/SnOpbDufP2aF8q0ptq9MqnUtJVe5t1c2WzVzj6Q+pwv8MJ5/u9bPpnqG7I1DU48/ hre3rVfgvPLmDXCh7dPxBScjn53tiMxlpNj1ZvEgqQ9Xu+gB4+OteVfUn0Mn6BlhkD2vo9fs1CGe tf0TeV8fzOIcrO4QS4WxJH4Vg+xtBjBfy6kJzEuNoMF6jH+uUhmGBzBFCbdujBi7d6059RDFxlrS +JmZMZaQNYdJfox+5NXAvIbGz5ES7WgIrnx5g4rccJG0JT75By9rWe6zfGN8y+tzjL+S714VpSVL 7ZCIPi7BerpTULpbdsIyyFKjTYqrfyaxdhtsFBWzlDa2TZ81VS2hWoy3/2R8BYryREZpuyFlMqKE LgRV4KgIqd99YKkk/JM4p2kehINSleKTOfI9/6TWlIH7Qb3lb8Dju03t0r7ByAVocOT4xLs+7lz/ KKX6EZCoADgXNMa9RCmr+JF7pPORYjzVrmgg8vAwefbNvjxpZmPGHiBSHsdD3oPWs4w2Q6AYBr5s LRDTJxOmN8hr/hHgYCjtpEGzmJ//3fOXcl00proWGg6/1XcUqTphWAElV56jMhrVJDOj2JW2NGYQ Rvi2EZ6W0urFzHV8xC0o5poPujKaq5e1wd9a5P5R2JtL6V76fb8zclIS0nZ6FAU/1pmK0OuaqOpZ BCjX7hYTSARUI7KvQnhdHmSa93575IPN+7Nwn2BNevItahM1Cgl8/pGdvxm69ovcwyusZGKcigPB JlyqXUiRkzOEe7bEoEzSU0Q4oj+/s7pIusb8PXaJm/3Pi1IJMtsFGjaScXbz9UA9E7Ss0AOkt+Td 9FkpUiVZtqiqn+2sd23sAMbsnspEk9m1b1ZLtxmPm51+MGIuJWweXrGpBo/wW/kgRw4VgfzajJZ6 2l/Tz8bBhTUunY+XJQFNqG+TykBdIfLg43UosxDZGpLn3MSf1KAKRQ7E0lFWd27cFmSPMbmbbCqp SKGx05HeWidK28sOesKFtc71TnHJ1nIJvyiPOvvhQgMqKlkLfRHf7+ftDIkk2DfRmIHcsWJwI15L Fppc5lMdn37SiJUDsW9Av8dvExgczMnVkukboiniMhvEqlqWvkz08ZyTPSLSrKmimEbEvVsSW8jV Iy0NRUgamKs9SHRRwWEytw3La0rrWL9EKaf1gIC6wuFDACQGF4z7sD4bQ4b5GjVd6okP8r9HjyBu H+s1PuYtNebJAjyJ1J5Hu8oVH68jUfTtkqL1ixe66dkf5eMfj4O1yrEV9Ow2uNovvCDxhdS3m8/c GFz1RLZJ+1jv/IiGl6KggNq6WRj0nq7ZX8lGXcsdD25S4JF8puh4SWWA4Q/D7tPs0PDBAIBHPT3O NAB8FY0zode+MenV9S0TCefMt7EmsqJ4XKA79kA3BJLgRSFauV652NO+mXvReBBF/V50i/5rHZLc dKOmvipZI+bF1UxXac+8WttmwM0PKmkXozajK/usgbOI8a9pffrsHtyUUX1fbszirSrpy0edP0zD GxQBV+LTck5c3Xl1K3s/lezKZJa4RUZ+8ewgaf5BU0WoUvM5sXOCmIDWuJViczxaDBW3mBGT+Qt1 gOOx8dvUpueFdwsJwBzT/RRUDEKkgigkS8EQtI5NcS+cKCUyxa2U3/w6uvzvqN3gq/+sHR28E+WW IBsWIWMQ8w7qQa2MMqpnxcHZV3tfwpcuVKK80LQ/k3nFdPv4CFgfXM2W8LxiirJm8Qr5HH3YgGB0 nffQvz4hVg891FzoJ+qPzGqTHP1oe5m9G47dPEh7/DpZ4u11XH3SlWD1rqxaoU3loRh3NQre0gt2 +LJWZ9CUSLfWqe+dyFysuZ8diHAd+rdmr9kTJe+NJq+v9caDX/59w5JGPS6i51/CcbDX+1zh2aIz JjdQV/EQ/+dE2CGHiHjGNqbP/fNCNyVzc5Ei7YnM6V1bx90jOUlk26d2iqlwcWh7Qn2e5nDgD2i4 mNXFj9Fbjhur4/h1FIORE+rY5PbeWYPEoaHugKp3BuPj7IKuE18PPLkKvSarrBgBNbGgOqB74pmG VutYdg7GfRyLsc4+VK9bnMxM0TlcTTHUY10itpiXCExSysyBtLFvOv73t8tCV5O9fGSGfI/LDBsW igN3YvzwSQ22O1ug+y+u4dg6BNdcGy/ZzNlOR0Mxh6fXcyq3mNZdVebOH/7av0oWUmNn/fKq6m6/ 3MfJpgmC3EsEP+qbyYSi4ijXyS3v08vonuTrEZradkl4W4R0nChHLyouO6r2892MmKTb0HZJ+YY4 ti/f1LqAf+cX7fQz6gmy5KHTmsSF7shhUs92Y771TcK6fSDyZQioU/FJM7WokokUj8SjJPR4kPCR n8f8zf0QWi6crbex/KtkDDS/2eldSHOqoVPjBDN+J2l4asSj+KEotFGXGVrCi6BW+CxKl6yEHY81 NbiYypWXNoZK+QVV6xpEEIF6uzRfppzGpyNoTuwSBgyLKlymsf8uuXHgyvoXKFm+D149tSUhll5b aVsq0lAYWlWfyNamzDeTr6pD62N3wnBR9hkQRQWQsdW67LkOxT63q/I2NNVspOSQLJpcSewMJQ5u X5Tmlxr6smLhFJ7so7tSOc2WjZHBYyxh5/2hwGCdEEqv0InSn9kSiZD1ZRbl1zXCmXLfFXm/hoKF sukipTMsdXM5AvrLHUNOgDXz184YwvBUBKuhXwkP70/P6CFgnhXt7K7ZmHBV/l2DuZNPu/CTzO+o MPBuQFZ1jhgLUcGOMDMqf9DhV06z723ioMGtnOLRn1q7edHOf789qlmbECWR460iiWY4kDnVcEcI IZVjURCnTGYwI+JqcZ5JSlhq2z0q9kmL8RaQjczxPM2HfbJu+twcQbeHYeQ+dioZfLRy7JkiqNdc N1r80LsBsrwMarEje8MxxUVREvvVzzxkLDBR1nHsEW+MzzOu0AQI/FcrXNvO8+PHaXbjIuBABnc4 0Yt27yevgOKSnYmRnVV6VtPRdPakSnQhJOznkO3IZEyw+x2rM2eOk+jHqfcHqY9Oa8behWfe2mNj FlVWXTQzH5l//hbdQMV3diFTV/sZrrx8kJ4gPB2DO191nqplsFvNofht9o5MbeUCyyChWzmNqaKk mkZEG3Xjxw7QgsZ2zK499ENyCiuKteOLaQteZ3R5K2wPqe0enVRGGoBG/eUs0EvwQavaMUaFJpNw Fhdok/XJpyyJflNBWJ+h2DbIvdc/t/NEbiSfWY5IhGzbgqqcwZDMeR56w6ulDBha4a+wT2NS+5iZ kwxZDC7dCPOsQ5DBoPAfODY7EEyPb3yL5AOlK0BZitpEoCAxSUr6HfoDI3u+0/yPbhKFRleip9tw 7VS/B8KUlYDYcULfOx62ckoJLPX5Xpk25SHtaAdJRxomx1uHWmdBxz//Noexj04oHp2NAxpQ5bve EIb179S2Dmtlj8zHELRpXI/fpMCBEpovuSnGZARib3e0i668Ykc7up1b95ekwRabNYCKg8CT4Rtw JkMFv4pS00OxaRgQMb2Mun1DjfMHYUmzyE1CVGNlcx7Pv00X2GMo20pL5B/4cRzduL/iwba0jIrd YnD+alAEsKgyTlkLxFd73NbNpUlSRYZlE4/5eRtcY25yTkr3pq0ZhqUlJrlLMmWryafOENLaMqrG lM0sUTpD4Ac38QU6dBX2Sl6QBRxmdznwyKJQm3TSiizJIuHDK87aSH3YmS0wWVEAKYZgNBMDBcv4 TDrsDxQV/S2WitH5XTIkJEYvmHQEIvCBkBxDu65REtXSLOQsf1IOsauZStIRkkc8awwUByW76r9z CX1MarlOxuZkU2tlqEywsH667WJNpSpuP0VgXnr1Jd00sZLepSdb5rA2mgQQkdKeLSCY7C9eU/7X /GL6u+gh8l+92vrasQJfQ0mvvXkER2VwqbCREsD4deVJ4+RDopJRvTP+2HMI4zHJOB8WDksqxgBV 2HubtFAztddwmJIMs/ADVekkSFvMsIlbNlb5SzM94157R2NBEtpKp68bgncrRip5KaKPMb8q0+WC FS5ttftpkmO90W/wyimYh6jpdCO1RhUKAyFJ6kIB9PxR/aZXxeG6KogxKmyr5/u6mC66tT8OeE/p hOMz/T1aIB+kgF9fo0OtmZ72aftAZgCbS4cH4XWRiiLEis0+ebntGRFVdtM12048tNpFFeyZExkQ kNnFBM000HPyp9kgwkbE8xz7bF5PsNFTeSA5WixJddDEq7E/8YGyLUIixme6SyLbsWlGz+AIoGm6 iPKvX6kSqFaLLaL76XFQWovT5g0j/H+QRGc9lW4b0lll+w+HTGpH4WxNGxW+NUevDuU+NR7to892 iQViaoElpwkF+/4ezeKDPnEC6rAxBV/DgHuXDmAfzrhnXdKqDMakJ65Awb4stZh1Krrm05DcaXSc e1FGEH288lGPm7hnR3BnfXJOBj6iR7qET5baiEeXUtJiNW/pj0VrTCuJudPYptEt/8euISDprNvx ThT1Yf5rwn/K+GplzKrMSkDvE4k2x8BXaHz+SeHemsuiXYs8NMxLHmdfa0p0KE6T0kAY5Nk43R7a /4VHuAXpneG+dYvKJqhd9b0I5CppPld5me5f2XOAZtWBrXyo5TY9M10GPBgINMTWFCxA4CVePmT0 br0tti0WfBQHHgoh9YX2i+5kRm5+FGktyiTKl6vwGvC5XmAsJ8duHMIYaz0sJPTxw3ux71YoQdPn RQdSNsv82w9WEevEM1DKP2cYKZbUXxfzpZcZbR9m8Y1hW+OySL1X3FSEJhxTpDXX6bUfviG/tW3s Q3LyVuWYnOytnBlB9hS5xLRplqzN0NJxnmHYlYD0Zv+H0/JdGojtHCgYq0vcDAtZOGQ1Q3em8jeL 5xExsHvpN5wKgBkpM4mEbiu3dHtTwbxDE6LH2ZGf5j0uIiju8Pz5jvA/a3N8SQj/uu+2U3H9N13M tb4yMfRfsp+0sdrfgkP6bcHnk14t26fOTtae5hlTR246dIqBfdxGQ6lo8f1zFa7Q+xgJI7hwW5Po HpEZUSYUq7qRkeAxt/sxzvieB9k8fgePYfRGnCUESxVRsghzYfmWTNhg8fHRDM2Qf1L9FcJonPrf hH4MeNxettXbtESed2scdi4g9M8Nv+xui2MlcwNVsGPUpc02D0E9y69WOTRUT8C+scPNXL0lwcLJ xJko3+5vGmwP9Iz5ROxVSJXN/ZmSigEDdzOnkGWGm+CnZ8T20soGjD6GFFRojjtFkswytbD4L+Pa teLF/DaIsSUg6SwmxuH84UaPMbJZm9S/XQEfuQK3HSahWWGBluCuyNJ8uepAALXn8L19Y/dh+T0f 4r8WNWOM4I1QxI6r6Xv2zjhSNmySiMpGo+wfgTgs3VWbt+ClxlPIZ+IRCmw1Wl/4nCDLS/KZhfmc R/No/4IMNWZUSTzmT2fVMWUSDxzp8PeE9J/tyrh7IoVgzFr3QiqtVFYWf6HOKDHcx7z/mB1QubgJ lrB/VhJwo65br//kzmxhJXLBiBEzgj1gq8C9u/t/4G82/7/U/wZ7+Oyw ------=_NextPart_000_0007_01C0C5B3.62DC5A80-- From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 15 11:13:30 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id LAA07232 for reliable_computing-outgoing; Sun, 15 Apr 2001 11:13:30 -0500 (CDT) Received: from mail4.lig.bellsouth.net (mail4.lig.bellsouth.net [207.203.120.43]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id LAA07227 for ; Sun, 15 Apr 2001 11:13:27 -0500 (CDT) Received: from u8174 (adsl-20-83-134.lft.bellsouth.net [66.20.83.134]) by mail4.lig.bellsouth.net (3.3.5alt/0.75.2) with SMTP id MAA08387 for ; Sun, 15 Apr 2001 12:13:25 -0400 (EDT) Message-Id: <2.2.32.20010415161409.00737064 [at] pop [dot] louisiana.edu> X-Sender: rbk5287 [at] pop [dot] louisiana.edu X-Mailer: Windows Eudora Pro Version 2.2 (32) Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Sun, 15 Apr 2001 11:14:09 -0500 To: reliable_computing [at] interval [dot] louisiana.edu From: "R. Baker Kearfott" Subject: New preprints available Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Colleagues, I have just made several more of my new preprints available (in Postscript, DVI, and PDF formats) at http://interval.louisiana.edu/preprints.html The new entries are: Verifying Topological Indices for Higher-Order Rank Deficiencies, submitted for the proceedings of the 2000 AMS-IMS-SIAM summer research conference on Algorithms and their Complexity for Nonlinear Problems, July 16 to July 20, 2000, Mt. Holyoke College, edited by Christopher Sikorski. R. B. Kearfott and H. Mu{\~n}oz, Slope Interval, Generalized Gradient, Semigradient, and Slant Derivative, submitted to the SIAM Journal on Optimization Existence Verification for Higher Degree Singular Zeros of Complex Nonlinear Systems, (joint with J. Dian), submitted to the SIAM J. Numer. Anal. Best regards, Baker P.S. I encourage others to also post announcements of their recent work to this list. --------------------------------------------------------------- R. Baker Kearfott, rbk [at] louisiana [dot] edu (337) 482-5346 (fax) (337) 482-5270 (work) (337) 981-9744 (home) URL: http://interval.louisiana.edu/kearfott.html Department of Mathematics, University of Louisiana at Lafayette Box 4-1010, Lafayette, LA 70504-1010, USA --------------------------------------------------------------- From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 15 12:04:16 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id MAA07557 for reliable_computing-outgoing; Sun, 15 Apr 2001 12:04:15 -0500 (CDT) Received: from clmboh1-smtp3.columbus.rr.com (clmboh1-smtp3.columbus.rr.com [65.24.0.112]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id MAA07552 for ; Sun, 15 Apr 2001 12:04:12 -0500 (CDT) Received: from oemcomputer (dhcp065-024-174-102.columbus.rr.com [65.24.174.102]) by clmboh1-smtp3.columbus.rr.com (8.11.2/8.11.2) with SMTP id f3FH1Hf20219; Sun, 15 Apr 2001 13:01:17 -0400 (EDT) Message-ID: <000e01c0c5cd$e1e32060$66ae1841 [at] columbus [dot] rr.com> From: "Ramon Moore" To: "R. Baker Kearfott" Cc: References: <2.2.32.20010415161409.00737064 [at] pop [dot] louisiana.edu> Subject: Re: New preprints available Date: Sun, 15 Apr 2001 13:02:40 -0400 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Thanks, Baker, for this valuable link to your outstanding and on-going work! Ramon Moore ----- Original Message ----- From: "R. Baker Kearfott" To: Sent: Sunday, April 15, 2001 12:14 PM Subject: New preprints available > Colleagues, > > I have just made several more of my new preprints available (in > Postscript, DVI, and PDF formats) at > > http://interval.louisiana.edu/preprints.html > > The new entries are: > > Verifying Topological Indices for Higher-Order Rank > Deficiencies, submitted for the proceedings of the 2000 > AMS-IMS-SIAM summer research conference on Algorithms and their > Complexity for Nonlinear Problems, July 16 to July 20, 2000, Mt. > Holyoke College, edited by Christopher Sikorski. > > R. B. Kearfott and H. Mu{\~n}oz, Slope Interval, Generalized > Gradient, Semigradient, and Slant Derivative, submitted to the > SIAM Journal on Optimization > > Existence Verification for Higher Degree Singular Zeros of > Complex Nonlinear Systems, (joint with J. Dian), submitted to > the SIAM J. Numer. Anal. > > Best regards, > > Baker > > P.S. I encourage others to also post announcements of their recent > work to this list. > > > --------------------------------------------------------------- > R. Baker Kearfott, rbk [at] louisiana [dot] edu (337) 482-5346 (fax) > (337) 482-5270 (work) (337) 981-9744 (home) > URL: http://interval.louisiana.edu/kearfott.html > Department of Mathematics, University of Louisiana at Lafayette > Box 4-1010, Lafayette, LA 70504-1010, USA > --------------------------------------------------------------- > From owner-reliable_computing [at] interval [dot] usl.edu Thu Apr 19 10:44:42 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id KAA12806 for reliable_computing-outgoing; Thu, 19 Apr 2001 10:44:41 -0500 (CDT) Received: from cadcam7a.ame.nd.edu (cadcam7a.ame.nd.edu [129.74.24.151]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id KAA12801 for ; Thu, 19 Apr 2001 10:44:37 -0500 (CDT) Received: from localhost (hliang@localhost) by cadcam7a.ame.nd.edu (8.10.1/8.10.1) with ESMTP id f3JFiVr00847 for ; Thu, 19 Apr 2001 10:44:31 -0500 (EST) X-Authentication-Warning: cadcam7a.ame.nd.edu: hliang owned process doing -bs Date: Thu, 19 Apr 2001 10:44:31 -0500 (EST) From: Hongkun Liang X-Sender: hliang [at] cadcam7a [dot] ame.nd.edu To: reliable_computing [at] interval [dot] louisiana.edu Subject: question about interval ODE Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Dear interval researchers, I am looking for reference to solve first order or second order ODE with interval variables. Any pointer to this kind of problems is welcome. Best regards, Hongkun Liang --------------------------------------------------- Ph.D Candinate Department of Chemical Engineering University of Notre Dame Notre Dame, IN 46556 Tel: 2196317893 Email: hliang [at] nd [dot] edu -------------------------------------------------- From owner-reliable_computing [at] interval [dot] usl.edu Thu Apr 19 11:30:02 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id LAA13164 for reliable_computing-outgoing; Thu, 19 Apr 2001 11:30:02 -0500 (CDT) Received: from sciences.univ-angers.fr (sciences.univ-angers.fr [193.49.146.182]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id LAA13159 for ; Thu, 19 Apr 2001 11:29:58 -0500 (CDT) Received: from elbouddi (pbastion.univ-angers.fr [193.49.146.252]) by sciences.univ-angers.fr (8.9.3/8.9.3) with SMTP id SAA17712; Thu, 19 Apr 2001 18:29:45 +0200 (CEST) Message-ID: <007e01c0c8ed$f14af640$5d5113ac [at] istia [dot] uang> From: "luc.jaulin" To: "Hongkun Liang" , References: Subject: Re: question about interval ODE Date: Thu, 19 Apr 2001 18:29:44 +0200 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.2615.200 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2615.200 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk > Dear interval researchers, > > I am looking for reference to solve first order or > second order ODE with interval variables. Any pointer > to this kind of problems is welcome. Dear Hongkun Liang Maybe you can have a look to the following paper and the references therein. L. Jaulin (2001). Nonlinear bounded-error state estimation of continuous-time systems. (Submitted to Automatica). Internal report : R-LISA-2001-01 http://www.istia.univ-angers.fr/~jaulin/voltera.zip Regards, Luc From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 22 18:48:24 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id SAA16881 for reliable_computing-outgoing; Sun, 22 Apr 2001 18:48:24 -0500 (CDT) Received: from cs.utep.edu (mail.cs.utep.edu [129.108.5.3]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id SAA16876 for ; Sun, 22 Apr 2001 18:48:19 -0500 (CDT) Received: from earth (earth [129.108.5.21]) by cs.utep.edu (8.11.3/8.11.3) with SMTP id f3MNmGt25417 for ; Sun, 22 Apr 2001 17:48:16 -0600 (MDT) Message-Id: <200104222348.f3MNmGt25417 [at] cs [dot] utep.edu> Date: Sun, 22 Apr 2001 17:48:15 -0600 (MDT) From: Vladik Kreinovich Reply-To: Vladik Kreinovich Subject: from NA Digest: a conference of interest to interval researchers To: reliable_computing [at] interval [dot] louisiana.edu MIME-Version: 1.0 Content-Type: TEXT/plain; charset=us-ascii Content-MD5: kEqD+8DYoAQLpkca4UH0AQ== X-Mailer: dtmail 1.3.0 @(#)CDE Version 1.4 SunOS 5.8 sun4u sparc Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk From: Shaun Forth Date: Thu, 19 Apr 2001 10:43:02 +0100 Subject: Workshop in Cranfield on Automatic Differention The next meeting in the series of Joint University of Hertfordshire/Cranfield University (RMCS Shrivenham) Automatic Differentiation Symposia will be held at Cranfield University RMCS Shrivenham Campus Swindon England on Wednesday 23rd May 2001 Coffee at 10:00 for a 10:30 start. Finish around 16:30. This is one of a 6 monthly series of 1-day workshops presenting theoretical developments in and applications of Automatic Differentiation (AD). The atmosphere is informal and newcomers to the field, particularly those involved in industrial applications, are most welcome. The timetable is still being put together and will include Mohamed Tadjouddine - Cranfield University (RMCS Shrivenham) Project on AD by source text translation: progress report. (We now have Fortran code that computes Jacobians twice as fast as ADIFOR, TAMC etc. on a non-trivial problem.) Chris Bischof - Technical University of Aachen, Germany Recent AD Projects at Aachen Shaun Forth - Cranfield University (RMCS Shrivenham) An efficient implementation of AD in MATLAB. John Pryce - Cranfield University (RMCS Shrivenham) Exception-free interval arithmetic on the extended reals: progress report. Please contact us ASAP if you wish to attend, and if you wish to be a speaker. If coming from a long distance, let us know if you wish us to arrange accommodation for you. Looking forward to seeing you there. John Pryce & Shaun Forth For further and updated details please see http://barrington.rmcs.cranfield.ac.uk/directories/deptdir/esd/research/amor/re search/workshop?view=www or contact Dr Shaun A Forth Applied Mathematics & Operational Research Engineering Systems Department Cranfield University, RMCS Shrivenham Swindon SN6 8LA, England tel: +44 (0)1793 785311 fax: +44 (0)1793 784196 email: S.A.Forth [at] rmcs [dot] cranfield.ac.uk From owner-reliable_computing [at] interval [dot] usl.edu Tue Apr 24 04:16:06 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id EAA18630 for reliable_computing-outgoing; Tue, 24 Apr 2001 04:16:06 -0500 (CDT) Received: from rztsun.rz.tu-harburg.de (rztsun.rz.tu-harburg.de [134.28.200.14]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id EAA18625 for ; Tue, 24 Apr 2001 04:16:01 -0500 (CDT) Received: from auriga (auriga.ti3.tu-harburg.de [134.28.20.118]) by rztsun.rz.tu-harburg.de (8.9.0/8.8.8) with SMTP id LAA23799 for ; Tue, 24 Apr 2001 11:15:59 +0200 (MET DST) Message-ID: <08d501c0cc9d$eb8cf830$76141c86 [at] ti3 [dot] tuharburg.de> From: "Siegfried M. Rump" To: "reliable_computing" Subject: conference Date: Tue, 24 Apr 2001 11:06:59 +0200 Organization: TU Hamburg-Harburg MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by interval.usl.edu id EAA18626 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Following an announcement from NA-digest which may be of interest. Best S.M. Rump *********************************************************** Date: Thu, 19 Apr 2001 10:43:02 +0100 Subject: Workshop in Cranfield on Automatic Differention The next meeting in the series of Joint University of Hertfordshire/Cranfield University (RMCS Shrivenham) Automatic Differentiation Symposia will be held at Cranfield University RMCS Shrivenham Campus Swindon England on Wednesday 23rd May 2001 Coffee at 10:00 for a 10:30 start. Finish around 16:30. This is one of a 6 monthly series of 1-day workshops presenting theoretical developments in and applications of Automatic Differentiation (AD). The atmosphere is informal and newcomers to the field, particularly those involved in industrial applications, are most welcome. The timetable is still being put together and will include Mohamed Tadjouddine - Cranfield University (RMCS Shrivenham) Project on AD by source text translation: progress report. (We now have Fortran code that computes Jacobians twice as fast as ADIFOR, TAMC etc. on a non-trivial problem.) Chris Bischof - Technical University of Aachen, Germany Recent AD Projects at Aachen Shaun Forth - Cranfield University (RMCS Shrivenham) An efficient implementation of AD in MATLAB. John Pryce - Cranfield University (RMCS Shrivenham) Exception-free interval arithmetic on the extended reals: progress report. Please contact us ASAP if you wish to attend, and if you wish to be a speaker. If coming from a long distance, let us know if you wish us to arrange accommodation for you. Looking forward to seeing you there. John Pryce & Shaun Forth For further and updated details please see http://barrington.rmcs.cranfield.ac.uk/directories/deptdir/esd/research/amor/research/workshop?view=www or contact Dr Shaun A Forth Applied Mathematics & Operational Research Engineering Systems Department Cranfield University, RMCS Shrivenham Swindon SN6 8LA, England tel: +44 (0)1793 785311 fax: +44 (0)1793 784196 email: S.A.Forth [at] rmcs [dot] cranfield.ac.uk ======================================== Prof. Dr. Siegfried M. Rump Inst. f. Informatik III Technical University Hamburg-Harburg Schwarzenbergstr. 95 21073 Hamburg Germany Tel. +49-40-42878-3027 Fax +49-40-42878-2489 http://www.ti3.tu-harburg.de/rump rump@tu-harburg.de From owner-reliable_computing [at] interval [dot] usl.edu Wed Apr 25 08:16:47 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id IAA00856 for reliable_computing-outgoing; Wed, 25 Apr 2001 08:16:47 -0500 (CDT) Received: from maebashi-it.ac.jp (zhong01.maebashi-it.ac.jp [202.236.152.193]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id IAA00851 for ; Wed, 25 Apr 2001 08:16:42 -0500 (CDT) Received: (from zhong@localhost) by maebashi-it.ac.jp (8.9.1/8.9.1) id WAA09763 for reliable_computing [at] interval [dot] usl.edu; Wed, 25 Apr 2001 22:28:46 +0900 (JST) (envelope-from zhong) Date: Wed, 25 Apr 2001 22:28:46 +0900 (JST) From: Ning Zhong Message-Id: <200104251328.WAA09763@maebashi-it.ac.jp> To: reliable_computing [at] interval [dot] usl.edu Subject: IEEE Data Mining 2001: Call for Tutorials Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk [Apologies if you receive this more than once] ---------------------------------------------------------------------- ICDM '01: The 2001 IEEE International Conference on Data Mining Sponsored by the IEEE Computer Society ---------------------------------------------------------------------- San Jose, California, USA November 29 - December 2, 2001 Home Page: http://kais.mines.edu/~xwu/icdm/icdm-01.html IEEE ICDM 2001: Call for Tutorials ********************************** The 2001 IEEE International Conference on Data Mining (ICDM '01) will include tutorials providing in-depth background on specific subjects in data mining. The recency of the data mining field, and the variety of disciplines that are represented, lead to many possibilities for good tutorials: * Short courses on areas of machine learning, databases, or statistics that may be "old hat" to specialists in that discipline, but are new to a majority of the conference attendees. (e.g., An Introduction to Hidden Markov Models). * Surveys of new and developing research areas in data mining. (e.g., Text Mining). * End-to-end descriptions of the practical application of data mining technology (i.e., applications that may be "typical" for a paper, but provide an example of issues faced in a data mining project that would generalize to problems faced by the conference attendees). * An in-depth coverage of a past research breakthrough that is now becoming a mature technology. The topics of interest fall within those described in the conference Call for Papers (http://kais.mines.edu/~xwu/icdm/icdm-01.html). Submission Details ================== The tutorial proposal should include the following: 1. Title and abstract of the tutorial; 2. Intended audience. Include prerequisite knowledge required of the attendees, and the expected areas of interest. (For example, a tutorial on statistics for people applying data mining tools vs. a tutorial on statistics for people building data mining tools); 3. Length of time needed (e.g., half day or full day); and 4. Short biographies of the presenters. Tutorial materials such as handouts and slides should be included if available, but are not required for submission. However, providing such materials will show depth and maturity of the tutorial, and will be a strong factor in the selection process. Please send a soft copy (preferred) of your proposal to clifton [at] computer [dot] org, or a hard copy to: Dr. Chris Clifton The MITRE Corporation M/S K308, 202 Burlington Rd, Bedford, MA 01730-1420, USA Important Dates =============== June 30, 2001: Tutorial submissions. July 31, 2001: Acceptance notices. August 31, 2001: Camera-ready copy of tutorial handouts. November 29, 2001: ICDM '01 tutorials. From owner-reliable_computing [at] interval [dot] usl.edu Wed Apr 25 08:23:55 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id IAA01048 for reliable_computing-outgoing; Wed, 25 Apr 2001 08:23:55 -0500 (CDT) Received: from feu.irisa.fr (feu.irisa.fr [131.254.60.80]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id IAA01041 for ; Wed, 25 Apr 2001 08:23:50 -0500 (CDT) Received: from irisa.fr (sigzag.irisa.fr [131.254.12.93]) by feu.irisa.fr (8.9.3/8.9.3) with ESMTP id PAA01599 for ; Wed, 25 Apr 2001 15:23:46 +0200 (MET DST) Message-ID: <3AE6CFE2.BC510B0D [at] irisa [dot] fr> Date: Wed, 25 Apr 2001 15:23:46 +0200 From: Sylvain Kerjean Organization: IRISA, Campus de Beaulieu, 35042 Rennes Cedex, FRANCE X-Mailer: Mozilla 4.76 [en] (X11; U; SunOS 5.7 sun4u) X-Accept-Language: fr-FR, en MIME-Version: 1.0 To: "reliable_computing [at] interval [dot] usl.edu" Subject: looking for some package Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Dear scientists, I am looking for a package of interval calculus which could deal with open, closed and clopen intervals. Thanks for your help. -- Sylvain Kerjean IRISA-INRIA, Campus de Beaulieu, 35042 Rennes cedex, France Tél: +33 (0) 2 99 84 75 99, Fax: +33 (0) 2 99 84 71 71 From owner-reliable_computing [at] interval [dot] usl.edu Wed Apr 25 13:17:39 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id NAA01701 for reliable_computing-outgoing; Wed, 25 Apr 2001 13:17:39 -0500 (CDT) Received: from marnier.ucs.louisiana.edu (root@[130.70.40.2]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id NAA01696 for ; Wed, 25 Apr 2001 13:17:36 -0500 (CDT) Received: from liberty (liberty.louisiana.edu [130.70.46.171]) by marnier.ucs.louisiana.edu (8.9.1/8.9.1/ucs-mx-host_1.3-virus) with SMTP id NAA14948; Wed, 25 Apr 2001 13:17:27 -0500 (CDT) Message-Id: <2.2.32.20010425181911.012cc374 [at] 130 [dot] 70.132.231> X-Sender: rbk5287 [at] 130 [dot] 70.132.231 X-Mailer: Windows Eudora Pro Version 2.2 (32) Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Date: Wed, 25 Apr 2001 13:19:11 -0500 To: Sylvain Kerjean , "reliable_computing [at] interval [dot] louisiana.edu" From: "R. Baker Kearfott" Subject: Re: closed and open interval package Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by interval.usl.edu id NAA01697 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Sylvain, What do you want to do with such a package? In particular, how will you be using the distinction between open and closed intervals? Best regards, R. Baker Kearfott At 03:23 PM 4/25/01 +0200, Sylvain Kerjean wrote: >Dear scientists, > >I am looking for a package of interval calculus which could deal with >open, closed and clopen intervals. >Thanks for your help. >-- >Sylvain Kerjean >IRISA-INRIA, Campus de Beaulieu, 35042 Rennes cedex, France >Tél: +33 (0) 2 99 84 75 99, Fax: +33 (0) 2 99 84 71 71 > > --------------------------------------------------------------- R. Baker Kearfott, rbk [at] louisiana [dot] edu (337) 482-5346 (fax) (337) 482-5270 (work) (337) 981-9744 (home) URL: http://interval.louisiana.edu/kearfott.html Department of Mathematics, University of Louisiana at Lafayette Box 4-1010, Lafayette, LA 70504-1010, USA --------------------------------------------------------------- From owner-reliable_computing [at] interval [dot] usl.edu Wed Apr 25 14:01:51 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id OAA02001 for reliable_computing-outgoing; Wed, 25 Apr 2001 14:01:51 -0500 (CDT) Received: from clavin2.unet.brandeis.edu (clavin2.unet.brandeis.edu [129.64.99.33]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id OAA01996 for ; Wed, 25 Apr 2001 14:01:46 -0500 (CDT) Received: from cs.brandeis.edu (IDENT:tim [at] odysseus [dot] cs.brandeis.edu [129.64.3.193]) by clavin2.unet.brandeis.edu (8.9.3/8.9.3/UNet2.1) with ESMTP id PAA26449; Wed, 25 Apr 2001 15:01:39 -0400 Message-ID: <3AE71F12.7C2B812F [at] cs [dot] brandeis.edu> Date: Wed, 25 Apr 2001 15:01:38 -0400 From: Tim Hickey Organization: Computer Science Dept./ Brandeis Univ. X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.17 i686) X-Accept-Language: en MIME-Version: 1.0 To: Sylvain Kerjean CC: "reliable_computing [at] interval [dot] usl.edu" Subject: Re: looking for some package References: <3AE6CFE2.BC510B0D [at] irisa [dot] fr> Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Hi Sylvain, If you don't mind using a Scheme interpreter implemented in Java, you can use the General Interval Arithmetic packages at http://silk.cs.brandeis.edu/interval/scheme/giaeval/README.html There is an applet interface at http://silk.cs.brandeis.edu/interval/scheme/giaeval/iaevaldemo.html The system manipulates finite sets of possibly unbounded intervals with the notation: [a,b] for closed intervals [a,b) and (a,b] for half-open intervals (a,b) for open intervals where a,b are fixed point decimals (where you set the scale) or are -inf or +inf for unbounded intervals. Sets of such intervals are represented by enclosing them in curly braces, e.g., {[1,2),(3,4]}/[1.1,1.2) = {(0.8333333333,1.8181818182) (2.5000000000,3.6363636364)} The only operations currently supported are the arithmetic operations +,-,*,/ ---Tim Hickey > > Sylvain Kerjean wrote: > > > > Dear scientists, > > > > I am looking for a package of interval calculus which could deal with > > open, closed and clopen intervals. > > Thanks for your help. > > -- > > Sylvain Kerjean > > IRISA-INRIA, Campus de Beaulieu, 35042 Rennes cedex, France > > Tél: +33 (0) 2 99 84 75 99, Fax: +33 (0) 2 99 84 71 71 From owner-reliable_computing [at] interval [dot] usl.edu Thu Apr 26 06:35:49 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id GAA03013 for reliable_computing-outgoing; Thu, 26 Apr 2001 06:35:48 -0500 (CDT) Received: from relay.kfupm.edu.sa (relay.kfupm.edu.sa [212.26.1.21]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id GAA03008 for ; Thu, 26 Apr 2001 06:35:38 -0500 (CDT) Received: from relayin.kfupm.edu.sa (relayin.kfupm.edu.sa [212.26.2.26]) by relay.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id OAA192154; Thu, 26 Apr 2001 14:23:11 +0300 Received: from iws20.dpc.kfupm.edu.sa (iws20.dpc.kfupm.edu.sa [196.15.32.20]) by relayin.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id OAA24870; Thu, 26 Apr 2001 14:21:26 +0300 Received: from itcH70.itc.kfupm.edu.sa (itc.itc.kfupm.edu.sa [196.15.32.8]) by iws20.dpc.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id OAA23144; Thu, 26 Apr 2001 14:14:48 +0400 Received: from smtp (webnews.kfupm.edu.sa [196.15.32.251] (may be forged)) by itcH70.itc.kfupm.edu.sa (AIX4.3/8.9.3/8.9.2) with SMTP id OAA88580; Thu, 26 Apr 2001 14:30:40 +0300 Received: from 196.1.65.140 by smtp (InterScan E-Mail VirusWall NT); Thu, 26 Apr 2001 14:32:05 +0300 Message-ID: <000601c0ce44$92906b10$8c4101c4@icssarfraz> From: "Dr. Muhammad Sarfraz" To: "Dr. Muhammad Sarfraz" Subject: GDAM_CGIM'2001 Date: Thu, 26 Apr 2001 14:32:22 +0300 MIME-Version: 1.0 Content-Type: multipart/related; type="multipart/alternative"; boundary="----=_NextPart_000_0000_01C0CE5D.B524C4A0" X-Priority: 1 X-MSMail-Priority: High X-Mailer: Microsoft Outlook Express 5.00.2919.6700 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk This is a multi-part message in MIME format. ------=_NextPart_000_0000_01C0CE5D.B524C4A0 Content-Type: multipart/alternative; boundary="----=_NextPart_001_0001_01C0CE5D.B5264B40" ------=_NextPart_001_0001_01C0CE5D.B5264B40 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Special Session on=20 GEOMETRIC DESIGN & Modeling The International Conference on COMPUTER GRAPHICS AND IMAGING=20 (CGIM 2001) AUGUST 13-16, 2001 Honolulu, Hawaii, USA. http://www.iasted.com/conferences/2001/hawaii/cgim.htm Aims and Scope Computer Aided (CA) techniques for Geometric Design (GD) are extremely = useful tools in a number of industrial settings. Specifically, the = application of geometry to current problems of designing and = manufacturing have grown at a large scale. In addition to its critical = importance in the traditional fields of automobile and aircraft = manufacturing, shipbuilding, and general product design, more recently, = the CAGD methods have also proven to be indispensable in a variety of = modern industries, including computer vision, robotics, medical imaging, = visualization, and even media. The purpose of the Geometric Design session in this conference = (CGIM'2001) is to focus on interdisciplinary methods and affiliate = research done among various disciplines with the goal of stimulating = views and providing a forum where researchers and practitioners can = discuss the latest developments linked to Computer Graphics and Imaging. = Papers presenting original research with the theme of "Geometric Design = & Modeling" are being sought. Suggested topics include (but are not = limited to):=20 *=20 * Curve and Surface Design * Visualization * Computer Graphics * Geometric Modeling=20 * Geometric Algorithms=20 * Manufacturing Geometry * Computational Geometry * Reverse Engineering * Solid Modeling * Computer Aided Design (CAD) * Computer Aided Manufacturing (CAM)=20 * Font Design * Multiresolution Methods * Geometric Transformations * Mesh Processing * Shape Abstraction and Modeling=20 * Shape Analysis and Description * Industrial Applications=20 All Submissions should be addressed to: Chair: GDAM-CGIM'2001 Dr. M. Sarfraz Department of Information and Computer Science, King Fahd University of Petroleum and Minerals, KFUPM # 1510, Dhahran 31261. Saudi Arabia. Tel: +966-03-860-2763 Fax: +966-03-860-2174 Email: sarfraz [at] kfupm [dot] edu.sa ------=_NextPart_001_0001_01C0CE5D.B5264B40 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable *******************************

Special Session on

GEOMETRIC DESIGN & Modeling

The International Conference on COMPUTER = GRAPHICS AND IMAGING

(CGIM 2001)

AUGUST 13-16, = 2001

Honolulu, Hawaii, = USA.

http://ww= w.iasted.com/conferences/2001/hawaii/cgim.htm

 

Aims and = Scope

Computer Aided (CA) techniques = for Geometric Design (GD) are extremely useful tools in a number of = industrial settings. Specifically, the application of geometry to current problems = of designing and manufacturing have grown at a large scale. In addition to = its critical importance in the traditional fields of automobile and aircraft manufacturing, shipbuilding, and general product design, more recently, = the CAGD methods have also proven to be indispensable in a variety of modern industries, including computer vision, robotics, medical imaging, visualization, and even media.

 

The purpose of the Geometric = Design session in this conference (CGIM’2001) is to focus on = interdisciplinary methods and affiliate research done among various disciplines with the goal of stimulating views and providing a forum where researchers and = practitioners can discuss the latest developments linked to Computer Graphics and Imaging. =

 

Papers presenting original = research with the theme of "Geometric Design & Modeling" are being = sought. Suggested topics include (but are not limited to): =

 


=DC    Curve and Surface Design

=DC    Visualization=

=DC    Computer Graphics

=DC    Geometric Modeling

=DC    Geometric Algorithms

=DC    Manufacturing Geometry

=DC    Computational Geometry

=DC    Reverse Engineering

=DC    Solid Modeling

=DC    Computer Aided Design (CAD)

=DC    Computer Aided Manufacturing (CAM)

=DC    Font Design

=DC    Multiresolution Methods

=DC    Geometric Transformations

=DC    Mesh Processing

=DC    Shape Abstraction and Modeling

=DC    Shape Analysis and Description

=DC    Industrial Applications


 

All Submissions should be = addressed to:

Chair: = GDAM-CGIM’2001

Dr. M. = Sarfraz

Department of Information and Computer = Science,

King Fahd University of Petroleum and = Minerals,

KFUPM # 1510, Dhahran 31261.

Saudi Arabia.

Tel:     = +966-03-860-2763

Fax:     = +966-03-860-2174

Email:    sarfraz [at] kfupm [dot] edu.sa

------=_NextPart_001_0001_01C0CE5D.B5264B40-- ------=_NextPart_000_0000_01C0CE5D.B524C4A0 Content-Type: image/png; name="image001.png" Content-Transfer-Encoding: base64 Content-ID: iVBORw0KGgoAAAANSUhEUgAAAlgAAABNCAIAAAD1gI2KAAAAAXNSR0IArs4c6QAAAAlwSFlzAAAO xAAADsQBlSsOGwAAVUNJREFUeF7tnQmcXlV5/2fJ7FkmyWSyDtkTQkASICxlURHBtdYdSxEUXKit /q3Lh49WpVr78S9V/2otloqyFHGptVRkK4IKlUiAJEDIQkIyJJBksk2Smcxkkpn5f8/9ve8zZ869 7zvvzDsTlt77uZnPveee5TnPfXN+93nOs5T29Dxe8iIc9f6YK1fNfK7jyO71a9aUlHY9mo+aytNK Fpf0NixcfFxNxdIlzV7VtpKSoy/CPNIhUw6kHEg58OJxoKW18sCBirZ91WtfGNvWNZx0jK7s6622 voYlV/es1XZtJbrwy2c2lOahpnl37xBoZehgFHWSf6xCBio95kCYgUDAD/puWbUa5Ptjd4bUc89w OKeb0p6lvWUr/QuuVz5aapXPKC8BFy9dcnL54tJXVWzJzra1kGmndVIOpBxIOfAy5UBHV+nzO2oB v32H2zdunzqisxAcGhAKigyN7NYvDOiJo9RwoWDx+GekHksg7IPAx1etMkgDz5ae1gvsiSbA79Xn j/7d/W1IhwaKlKvQrxb0cMqSJZ6MmMLhiP7vSDtPOZBy4FhzAOHvheYxgN+OPVMHFP4AMNDrUGvH gDULmYZ6S6yJFJiIjrmkQx8FfRwNpEzGCkTPQBwcRhRkrGMDhH0QeMMPVv/4AcfPaz+TAT+QD0hD tuNAOgQUl85wHx0rt1UI9qhgJdS0l2Hw+fjjq9Tnn7+25IorT87CYYqFhfzC0zopB1IOvHQ5YMJf 4ZpPX4YDPMBCTU+I6Cs8C8RI6zAPmzRKAJZxbAt6CARKnvpNAnxVW/APKB1eFDw2QOhQEEWoQeA/ fmrJxEVd7AgK1bTtZ5in2SIOBizzVaY8EnyqDoiIOIiU+ZlrXQkQe8Xl47PNW1+6v/GUspQDKQdS DiRxQPi3dUd3IcJf0IFw6/ipY63ch0O/cgCEiRhZOApaz3HZMRER4yhID/5OpDosfN9xyD8lkHWk JcJ6iPvhD2d9+huruAACl57SCyKyzyeNKGYvhoiGalyYCtTmpv3CQCIMmiMOGhz+5tZUNBzyDyNt mHIg5cCLwIGjPUe3bBtXjOVLsKXnz0HAYwIi1z4Q0vDQzmefP7ST8qnTz1I/VjK9dnLt5DmBycyA DBKGxS1c4hAIVT585hIlh10QhDypakcOCEchhTPAX/1VK3pLpDSENgmFKDBff8kJe9ZWotI08APS KOFWgCeJUBak0ppKIhRABgLlqxdmeqPzyz/jzGc0kCcatg74ztIKKQdSDqQceLE4oP2/wvWfuejU Zl4uIAmw0IBQKPjgY/fs3fxUdXXV7JPOrJvUxBDPrl11aPszGqt26vw5i5acvPT8xM08oyfRuNR2 +xKlwES1Kh0GAw0vCgYGOyMEhPVM44kjs16zxEEdwtnKx0sRCrWHR4mASqDFbWA74ytLxV/QUZ4V MhmVNAmyWlfW8+suWU3hj35wAn8/cOXT/N27xhETHa32ttKLlAMpB1IOvOgcMBXoMBp/5sLCRAUp WGgoaJjX2XkYOKyprek41MG1cckwEgHRChEi23dttVsQdP7sswLFbC4+++IpdfJsMQ4vCooeHwtH Aggd8LApKEz67aol13/SGbMgn739gqVf/Jq7FgTyYv771qe5nXteCY4TlJjZp5wr/MN/hNPFg38s 2fT7jHUM1UzW/Kd/qr/hxn1sFkoGFQ2pmvRF/w+fEpByIOWAz4HhEgHjXDULz8DwJEAda2iyICVA ncp9LOQWUOSvcNEwUuUUqr4a8hTZcXLDRBARvETVp/JA+WmFAf2JWCgUHAkbGcPCYQfCeigWCiKl AUtSjQJFlBsscf3hf1ltSCaQo5WJhjwKDsDSBEGrb/gnQdPv3wiYsLiVRykWxv/DpCUpB1IOHHsO bHyuDiuYYRQB/SkE4qDvgbBu+4H4ZH0U5GnX4cNdXV2VHFVVAkWDQLuwTg7sdx1aTbBQzUePGaO2 0qYiIBq8GRjLtSOR+bmwMD8QCtIGJTiOnEQYoqBACOWk4BBjmQ9+cIuuzdXBNygt/EcZby7cNXEQ w1FGD8A49awonMNpzZQDKQeGkQMjoQWNkwcQ5lJL+iBEQ6wU0WpuWn7Hzu3OQGbsuLEA24GD+931 mHHqmULwjwqCt/ET6k1ZCtRRn3Kq8UgQ2Hm4A5WqcJRyyY5T5p68YOZp2No4aMzin39BeS4dqYTa AuFtsFg4QkDYDwUR0fILhWASECi5cMgHYuL1H3EKVdPEmugpCMyBhQ6e0yPlQMqBlAPHgAOCwBVr Ggv02xsySfHdwbilqGCytWv93Xc9uGPTaoBt967dlDRMakCkazt4UMAmGgzthHAS9ST58VQoCOrx FwTt7OiorqmhmtpSLhxlCP6WjpuBvlSIKJnPt5Gx6wCtAcJCUHDAUDW5OhkJ1Wi9aUQTUdC35Ayk QO3nsfN347UD/wZAuC9fveSX962U1yCHwaFvoeqrRg0UqZy1nUmxcGBWpzVSDqQcKIYDxwwCRWSe +C+qYLuGbEL9/v4b9u11yyDgJxlOUp0wTCKdnra27u08fLgaZWlNDcKiJD9ji5BPJYaCum1oaLRN R1en8zAypQREbk1G5NrfQTRQNGpzOdFLWTogCvpvMEDEYQfCegaz7UBhnjSi2HkitKlEG3XSXoo4 rGa+eXlfW0oAxe//sVTbh7TlApzT34+e0auGhnx+V3KWSARjyYUf/tYSrFgFitHg7keQHikHUg6k HBh2DsgjcM3zM1taknfChn3E/J7vvuGMxMHNTy73N/mk55R4Z9uEXAOEPuyJbIS/zEUkGoKRNh3A 0ioIOA1TJU26+p7KlFsfEa2fuJOGj2EGYIUAoV9nRIHQ4YowCaiTSzsXBn4+CgosOfxwaMuuapWC NLCpsYbSoOIUAXdkEaPKUopazBoBJI8kDkowNfNRu6Cft531QtRH67D/HNMOUw6kHPhfzoERNYdJ 5K0FhUmMCOqLVs/t2Xrbf969dsXvhHAAFQpMX9tJoZScM2Y0SVlq0p6PiNRBTOzDv6oq/1boaHJk dOs0qyZ0UqJNR1SmXJuVqT87OchxGHr5+Ee53QZ2QFgABVyac0JNfeVCt+J3rR9fVccFIVuffbrD fD+GxWo0oxQ1pInDIU4U3Wt6bUfQE8sycqQBm0BR4p1sbWi74taMIhQp07YDrQkXhq8qFBbKSYML BXXDrdCH58hwJgXC/+VLVjr9lAPDyYGRc4rIQ2UhKCgsEQoiC7IviMLT4VNWgLML4Rkw1tg4hQuk Q6lGB8UmyYgBUloP9fUToqdOrDTDHLM+lWzKniV+/ShRwTCqgV5Al/Y1TblKuZBs5+49vfu3cev7 cvgWsBNmnwjWWjXVlHGQpNvigbDecAiY8SUw86PngkxJV3x8ldnFmG7zb25s1b6gpDRTdWonz/d8 8AHSsFCA6kOjeG1YCATijG8SqvS0IjIbjzTFwkH9wtPKKQdSDiRw4JhtBwJ7fkQYIyXIlBQPyxLI ghLpJLclvtH6+nrEOAmIgwXC/D8RX5Waq6ZoAzIlR06e6lz4tzZvFX4D0trUNCQbYMSaGlPnJtYs K/5HDQ6h7QTJUIqCTHJjB2nomWgyLr7okmZ86oWClGvPD30msCcUpBAUJBINu4O65W/gU3/VsiUU AmmU06GgjnGBUrrimqEp5C8HBAC99AMB0EMhddCUMi4xbiBJ/RQ/97SHlAMpB1IOIAguX1X1wMoR sQsF+fwTbtttLhSMvxG0iI+t3I0sKLtQWbUkApIziolOZDXEMv6CRoDilMlTrb4qxEcpBOFoBcjp 9K+txLqlZMfO7Rs3bmhp2fHMhg2cQsGSnv0tO9Zva36Cv62trYX8/ITleeC8SCCshwgwBmzjAwR0 UdhrCgWHIBCOg2CV9gUBS+Swv32/i39GiSxfUGBKOHv0FgeQahuf22mXZsrUP1gInnEBlKpzJWCS Zz0H0Eu3BoHUoaGgkcDfVMBINaroppAeKQdSDqQcGAIHEARXrZ74m4en4yAvfBrwiAOb3yR4WkiH fnMJgnFLE/bG1q64Zds2J1HpdK4OkV7UPySEIW/Nm78AIQy3BwQvLGjkHUh5AIE9Rx2sBj0MyAFV SBRGDSMNj0WSNKgOyXqcp6OOQ9kswXmQ1cDPtLVW2b8oEgid3AbGIK6BK5LGzGSGa2CJCr7YRyHC nzBMh8KNcigGN514+XX7WEoOekmKkho5hGc6JHe6wiXNCjQKVWS9gAAukAJpC2FALJgtRISqVCjs 4296lXIg5cAgOSBBsPhI2YMcNqw+YBZAmYn+4bGnfPWgE5Kyxp/qUfuCqCIBP/n/yb9CPhUoIbnI bCVmpauyURmPQyoAira/GJ8RT+OoWcjEGTET6aYKJ8WqkrJxnUeqdWp0H+oSZT4fJnONWCwQZuSz UxyunHJKPzgEbIAlDGQY29lwZnIEOs76GEYFAEmAymEiXZxiSYpOvRnVV886sh26zgFapEwO5X7i WgpSIBBExAcDKFVX3BbyJtI6KQdSDqQcCDiAaagEQdux42KkXeYT34Ic7xITO1COReVNt/znivvv EOxJ4OOMFJ7uROcpzaesY6Q7laP97t0tupVbRUYz2Z8IIVxb5/5DR3sEpVKl6rC6tXUNPmrKKzHX j0pPOekKZ0TsWkUDt5HPfj/FrOljA9ucXPrbxEGLAUI3SYW3BlGcJHdKL7GwgUOhI8iHTCZtZxbe WqWKVAUdVECJKlQDq0wcZJMvoJhHtgVIEzkXcnJomzDqnCFKLr/EwZ4GsgQXQCBCp0EgXV23QmQ4 ktIj5UDKgZQDhXBA6tCh7QjGVZ2BLlRo6p8BSYlYCxZygoXAHn91IgiyAN5z53UbVtyPzYsg0G37 RbiC5OeEv8pK/uIp4WxSuroUYo0Dx4msmYyDTycLVjlYMmIAuXnzFvAX/APhRlePmzFl5oymmeY1 CIDRrbBQKGu7jBRyG+0+Yo/j+SBGCBcB6ASGkz5WxCiWDUTSrdPQZuHQ9KthPzU1VIMef2szz8st BggzqkVyBAIw5EUCDl1c7FOcElJqTCENMpmv7USYM6OYjfc4SxYO4SV/2VCUwIe7hdHN7e0PT7O9 RpXT7Yrr6uU46PSffcYvbehRgT3KNRA1TQrklkwXDhFPWSLhspCfflon5UDKgZQDcEDq0BVb+vK/ H2O25Nk1lGgoUFy98n4EwXvu+KfnnnjYmb00NErBCKjIj955F8hTIgqQrWvn/15ZyTWCoIOirNDm 2kaYRIVIUKunQ7wA+Ttv5vHHNc0EFIM4MnTCWNTnqWK28VfConnZ+5Y4kWDqoHr+ggWG0Ean1LOZ 2wiVRZsQ0QFqdKE6Eg0FyTaoLyDGhcVi3CdchGtSxiPtIc8pfAxwCC6iigxc4z2nvXogTZkCLSOE lRT+k/Kc4jNOhNkgNa1RJ/X8k/eFueFzQdYnUQgQimx2NzHnSR0KC+d8WjPlwP9aDoCCazaUFZM4 YrDGL4NltXLKk1AX/JO/vPBAGs744QdX88PKWBw1iyCqpxZNO96VMjTFA5YqVLfQl0PyqB/IRt1m kDh6akmdArKNQnVlnv4OvKPAb+YTIv8K6pgLf2JYADM6HbJE6JAGFCSJrgRBHWSfAmNk8CKzzP7G L6StL/ndeoeCfrnZOCFHgmfSagYHhTyVoMmRbdLKNdjGX7Sy0eEIE7CpsgO/SKfKuE5sfbQUmVUW NJAtC530SDmQciDlQH4OgILaFBwyo0YOBV0yh53PrnrkZ4iAv/2p04VCpJMCI4ARnDicONwhk1Gh QiIKuvKqKhdfOxs+G3ShCUjDwa6hpm+Spa4VuVupmuyQrOln91UYbquQCW0aEcmgwkgOX0j1OzS5 UMSLwoz6VMFOI2kVjShdZTqJRElhZB9hkd41UvY63SznkCXCjBu7PNZJH39izxKJWciFaB0Rs+QC r9RLWZGrHmdBedZ75fVsJSJEAnWoOo1Wc5P3hT+eWmTtyPqmFeTzU0/4oqdfrkyHIk9/BZ+oTLOD 0tXL/iDC4aNbN67p3PXLTY/7k3n73FMum3/WqLJRbT2HfvrMY3q0uHrSmTOPzz/nzdtfuP/ApqA3 mtDh+WPnzp46LVdzNcz19L3zTx1dVmtPlzevg+ZclRmoZnT1r19YM7TXo2nmpyd/z2+etnjKmAk3 rH+wcAIY9MSm4/w5WtsRpSR4p/7rTiT+ioXn5n+D8R+SvXf/rdkLLX52Rk/wY/7bxRcF7zH/L7Dw l1VITUxjik8fMRJACASSUGlD86OEXMHf3NJEMCk/nhm3vqkLSkhlh7AkghKqqCZccc37Q0hGFZnN siQBTuFgJN41zWyiRDmbuFCqChMEJZ+ZdGhxTU1A9DMgSoL0Yc8P6u2T6sdH9VNkaAoOoQ25+yfH yFSI4BOMLAoIUS3iFCh9I935alJu+2eHb40GTgStTBLBbNy1TE1q98+pmym3dIZR7GwKHbgSTZv6 Aej6PXBtEAhOG3izi6lETi937Sir0t+vueee55/UC048Hrzw0+DNR39/k55ymwsIdxzcC/CwCObv 8KLpJ/3g3EsBifhwb7nz23nabnjLlw1EGavpJ5/NQ/bWi79+5YO35KckT/Pvn3cZy31+evI051HH Fd/n8+Lce/8xf7X4Uw0dlI8oJbyRO970CRsR8LbXXSB5hbx6vXd7a/6gRc6ObzXoBAJveubhOOXw 0/9N8sNI/O0N9jUNWF+y4HBZhA4KDhl0wPqYw0gRapmSmJHihyk0jCYoJad28vzM8haKjDo+9hhb TPFIiR8RzeebxC+zatEjP6OFpTOU7GgJLgytLZ2vYbP1r6m5/rMXPm02iqDdp0poGs+VYVFPLc7c kFWjzloX1aJUlOy6gYJcqBCDT9ny+lagos9sfH2jULVFy+ofucxYBLre4UxjZHEDvHnlrVyrXEaq xFfT6FSD4K5HSwhA2r+rl+Udix2rD8v0gFBBHVtcWLwSUZB1kN5Y46g5YIdUoCZNAsZBUv62O7v6 kmUDcnn4ztq3ZW/LgJTk6oFpAkUD0pOfAFZnPjKG8OOAh3yg+A2PJSW8lzwoCFWIcT5thb96vXdr i6ym6+JnRydwrOaGjyZS7v8m+WEcMxR8eO284UJBMSqPUWi/pS3v+iT3QdShezc/BQZgoolFaB8g ZeU561CuCELBSWMq+QsmoTVNBL8g44TdyqzUT9LErQ2q3pwPfnQGIOfrSGklnS1RY+jc6TkjwkSt 0MssdFQoGuKpMGyCRpVtMVo0AKuTQb7qjBrWogoMHQj3rO0LoiBEAaJUCDIhrXNhe4fR1h1niSqA T9SJiHOFOhaXCOEyNWNu9ZlyMMyaRJVHa2j+oqH1enA9m6gK7ooYt6kZ4S4HKE5hLv8bb5SX6KVW rkIQKz6B125zsfv8g89wFjIWuMGiThzJ4qrUXBxkCvmHQ9c3NBDSiEyTeeXHg/xvF/1bMeu7r/I9 xpTk/8LgE8HX3A7t1Yt1syY08rf42dEJZBQoefPDOAb/LfGUwDpmeFMpBZga95ewEiaYSxxkR5A1 FhQkahoQ4ktCgpMAMCQsopkEJHCTf27nfmApY/+StQj1VaO2/WZ6SOO22yqMLE6VyJByBC8wlRK5 ZFCip0JNS+dLNTN+sbGcR0e0lwlhqFWdDBeLfSq5Vrt6tnPZR0+URlFQqt1KNzQCcWSeE+x3ZgIC RNmg/H3QoQOh6BCKmFxoMlkWkzLO7/KI4HyqbJUaWokv+fmFVo5Vi1+uIRDv/EINh5AX9GBkCHeB YcGtLwv6iH4M/msN1xC5Vq7yh1t7b2zmLLtmHX+5jY9I4amzXFISOwCkP7v7e4mAYR3SWyLxIBl7 UfaI6/v+/cFcQ6uawQPSnqj1e/ZHnFw5Vr3FR/erqUIwKBWaRk3e3X5ANePEWyv/wq9G+ZT68a2/ ez6RAGrmapjIqBGlRCP6XxWzf9MVfwtGME/VhPelz6nE34l+SPotJU7q7B+VSzIrfnbIgj4Zemsa PRiaR0LfkT7Wr51QjHVM8eTFJVEZxTz42D1k1iWbkkk/AICWfhAOLDHYSEQO5yYYgYG/I5ixoIli qmWOrENCgECKMqMYNLJzkVoSoAWNMrGwY0ntDfwyLoyR1AgNQk028+hWPYtvufS0AVdVzfY7oYT+ uTUBUdM0hggyg06Gvkco34nXX+L2CM1GRjhHrl1LK1H8T6H4HrT7CElY9NAbRj0B2S87D4rEvR+3 cNyz48j2PkwS6yovn9N9Vr3PRhav3/7hX6wEA4cFd3wxvtbEeysrKyu9zvEwOPztxp/98IGLr/xU rppqaJtnf/2pa7/3rdsqptb2XNNnttP9kYylzz98/rNnnrvw/DdckTgL1sdgskE/J32n6/EnbxY9 wSM6hF1dN4Z5y8rOmlh6eZ9r6dd6L/jUhy8+5aT3r1rzVPAo3kN8COr4u6EjR4n/Oo586Afc8k7n N/1pT0+P//bBlZ6H9/AUUpvX/RcAxgdQ4v6rg70/7qO533MiD++++POvudApaYqc3Ueu/NPxN3zc huPtHL15ixEQ/Ia/8twFV3/l4uJXhvw94DX/0vEX3NLmzDWP7NqEa8Sh7c+Y+4FsTLSvZteWdJ4m ft55dxsJT77e0vdJsEdWUxc+o2yDLSNQZsHMN5yR54bZiEopKnEQ/S0yayZ4W1aY83fy8qS8Dx6Z BY1vmBNsEzJisHOpCvoOEHPop1iJcKR/i8Pef/99xGHv/lh0mIiCrFws63EUhCDwzCeLJeaaaz5q JXyGx1GQ3liG4r2xMCVKBr4C8Gvf+pHr/Axsegc4WKyv+/ZPqVR6kYvtpMP6Z9m9/LMXXnPN9ynk OsByZhEnz++HCld/8gPo6774D9cHQ2QG6s8WCh14eyhID3956dtY30FB14P3SD3AIn+G3bMSQkaZ TdBwURIQmYvF3/zHW3lZPt+YjlCQ4+8++FegIPyPa8Ld59RVq6gZoCCtYHigYDhz3UShYPGzM2Pm DG/X748ToEfQwA9jgN9W0Y8xEyWIaNHdFNUBwh/nM5sfRgRc//tfcD5657/tWPcYKCghjN6d+BWp E31ElEgEQAaijxb9jG6zP7zRlW9sIvAzGa5PRozEKTpXhBoT6USJk0ojdSiIZSgI5plSVDRTAij6 PoK+xCZKJMCZm4dBYJ/tT9YIyNSzNBTIueBwu3ZL6vVRUPuXcvxw25lZHSydD10ixOchLloBM4iJ Er+IPqo08f5vQfnrg3KlmFdeX7+y7E7lEW/laErla6GYMjoSu/XLIQlzU9syfPlKhIkoKNmIJfKi 1513/jtPN7XnDTf8x60//3XwIX/hv4/99T3fFN9AwfiWjMkNSxafePaFJ5272FkxPbt554233b7h 2fWJco9JhEgYv/+5e2uXdN9qb4cOAxSRnSEL8Yq71m89uvPq0vtU2YH09HfNme32Lxtn1J9x7vG/ um051zdPeSLYSkSo/csrQ5ng/Q997+j6TIyoc3rnIPWKHn8IG+irZ74Lxan/e6Pa55f/u5XcfM7H 3vPB1/723pUt21oTe6CCKvP01u/e+eTHw9QDEihVZ7go8VkBu4LvA0dMZEsJfgdvAfmYLwMR89b3 nbn/aEfcWNfksBnTJr/tvefz6nkL1H9sy/pvf/mmbS/s9CUzKps4WOTs3nH5uRjI+O+CzmHvsjcu bH5y5/s/+LkXZh61p//3siuMq36TYbxma5DwMS+WUlQZB7c//zDCn/LNSn7SYcu3gRDYg/hFsgjf b92Uln4TAaG87P2tOHNCMDS1VvbICJASEiABCyn0/RwkUCbKo9TUU/VsUqbpZl15RJ5FVvNvTYT1 /Sissp4KeoWyotY2OCXFupKsAB3UccbKxRyCKN8aEzsmSqKtOLdvF9i8rHyc70dXITHFhF+IU4QI Y4h+Edoed+UulluUbkKH3OIxhAm6NXf5qLdV+PvLnT9u6VMME45Z22ATReMKBS9595u//P+uChz7 +FqvnzHWSV3eBs8XfvxdNWTxyoWCQOA3v/FJfezb8YnPve3s0z/yxNjtGVsj7xE+c7pjCQY8nMtd FDOPw62tORgEtbM/OI0NqpLnMzVOWFF79Q/7wRu9gZeX3NEHq+rzmmsyGjnrm2q9X9jRk9UMX3P3 140efwjVf3f7ovhi+o3rf2LK0gVzFr7jB87zASYg7gTLNOUfGn8GtBknH5ix88n+visQefaXlhh5 4kzxlFxyg/eFcc8OLMDiWMigjOUMVu/tewtXf9Lhuu6Z0ZUPOEHZP4SC06ZM+scvfxZkkieDjteU LH3HW887/rT3uO+M7KfG60afZL+QImeHs0RADJMC8s8bfTJDAMnoz1UBhEZMDyoP++3zO2o3bj8W e5A+5bKLAQIRAcm6jl+gNIrBCm5NJAVyK/FL5QYw8p2wyoZ8CVuGnmioscyfXb2ZkGcuEGYmoxEN kATG6sSidZtfBIXmGmGOENa5IaJJgSK+b+cStMuSapUNXNV5QLxJtMxILoxWzTBSdYauGo3bmyAL UigzFiCH3rnOHq1y1FOaCC80qCvUkTVsydT0Y41GzzPlMR+JNp7JEDRrd6qafeUIgvQmYjCo4Vad gN8UxnN3GUkvqYtcuAWRP/nBN2756VcS3dsv+fhbUTE5TVd0XjS6z2siblXoFF9/3MfO3COrfxig IKPU1NQgUoxa6BKDBYdvfxhYD6KYLd/SF9MhaMhibaIeo4O+8c7/euXPg0JWyU0z24BbnXqKMtCU peeduczoT7T53HdZgzXnApo7OjoQemygL3/uw4YE8WUaUpd8+Ez1IG+TQGClAqAeOKgMLyWBltIo x/5I1/5XDjpMQ0EeMaO4sS4oiEbhkbW3UtNHQfXGr+uCE0+3HxIXX7j+w/57KWZ2q653cn9wyEkj iGPwiS9exu9wRP9jIg7iOz+iQ9C5Ym03NtZY0G1pQTeu+hmCILag0ij6cpupQOWcQCf+rhu35r3O tQ94GTzIWsHwNGPh4kGL1IZxDwQqKyaLoSB4IyWqBZRBxJThjKsMckcHdaQ+NeeKzLZcFswy7hBR NigpbH148/nvnkqey6pM9dR2NOW2wV/DQqmOOaQlNr9JNbTebMShAyH44XvmSetooKKLAqNaqy0Q 5R/x7BN6GnM6HI20pxROQt/s0Vcujavv9eiYeFo/ubDf2C/Jm0TcAmPu+dF3/TUuoB2rSytBd2qL V+KyxVL44+uvxQwhvg6qExRlcfkDBaA/qA8bLNZsR+VhZ7BYx9HXR0q/H8wLdVrh9NomIFznj37x JStPtIdkkbUecPZgvn98cN1Jixe98fWv4US89oWneA8wwZrHEUX7tXFQL5ISzDv9HhBJj6vI6ULg 40ewK4zoHKcEvcL73vnG2+/K56I+ob7OuAqLApgf8uz4BLn+J7fn+pHQ7ebXVV77/c/ytcf5iSvf NdL/O7EUHV6vwUSC8QI8fupY1kn+UgEUJDrMpuV3AIGoQ+UC0QdgWetNwxXrkxVflZ0o5glMggG3 0EWQoKWf5ghtwd6eAYb1aZaolDjRzcWzrkGicrdZqxxd8NQ5/EX6Rm7xf3BoFO3e+dHUpEq1fUHh mTpXKihtB0K/maf6Np9ihaEXt747IzVdVopJDYS28fdErQcfBW1qAa+GDoTiGjpGYAxhzkAIvGEb DxWlQoYGOf8Sy+UamBUfYbfjeEwizJRr3Kzw566tZiDb+T34iZ+QSkFTepBcmKiktd/ES+QiEQ8Q tkDBOHj4NBOczOABkNPiFSypqs/yTYU8mEod9oriDLnkfX2WC/HFGrPPeBOBR57F2poM6ERI6DhV Br/tNOG4kLhoBEmhOWxk61Qn4rUREBcH8/wkgBwQBWnpY598X/BeiqfENydhoL/59CXTpyUILjJc 8mHJTFpEeVzCprf3nPPaH932pVwfQGr41ovOt98SenifD8XMjk8QRPlc7hmMwq8FzTO/zERRdXj/ hyIOjrSNjARBFitlTWJpkjsEulDNBUiTglGClwrNWTCjXYy2xFj9hU/aHbTMSmbqAmD4NimCzAzK Zi02AYy+yJyRg2Dm9GKy0LOoojmdW9wy4VbGYCcSBE3nCZ3UZyxzkxdqCvDcjLKJlkS2sNDw2xeF xQEJmgZgdiG1p41iFxYrQMzh9AFeHdqPZ+hACH6AefhLKEqLQEiwp0jWSoSUjWrt3pa0o2RBCsql L1UOJkxaiKzGKUsZDi64pRzzHMvf65vVCOT6R7Fx2xt+1G8XaztKFKU+AV1Fw8m6/DvCXspHHA9Y vG75ylfyoyAzYvPG4MFALrDQoxq9XXXmhflREP2hWbUYr2joB/iIL9bU7N7ZGedtoEF93fPHFS4O Wm8Y3eRZuwtx8aaHhrqcxoGF9GDEAIEy3EVa+ta1/XS8hfSTn5L4FwZgXzemT0TzORx8YXzttj5i Er+o0OJef+Pn8qMg/fPziH9qUF7k7LBFohO+HvJjYTyA0Uj8h2V3cETFQW0EkibpV3+455c//ibR 0dauuKW7Za2MYnTI1lH4xK1gydZxFn2JdIYxKFHNGUA9ZGTB/lFDLQQ29ft0g5EwFw+ErX7c7iNe EKReykZQs3I3SiSnaqw+WIo89+VWLyw0khSezXS2Zv8JMdp0zHjxRwpbaUr7IrpFCGlNpIx1vUXE C6QDkyKLLS5m+gauMDDwZRwyELrfLtDittxOcdtvspcBFzET1RagsKp/pkDHk1cvPMEvR7Fp4hqG plJyxg/KMRY1dMxWqOdCSQeFu9mQoU4vSm8cGIgKPqEnyJvoCk9xRL7Ej8TFC3OP/LjFpFgQ48tH rmXr63+fT4dJb3jcB4ySaZ8VBjG9sP9ksa5eWJNoBx9oUIMNJ/UZh//ABf6ch8P4OD6FiXt7QQ/j b9qdBwDiPeRy8WZcGe4iMyFQBn0WT8nFd/+rTQ22f/4a51uJrtKsZP2J++IgXxi+DjPxi+qf7/hc /o033mwema/I2WEaCt+gHyzkY2LAvc8R/d+6dUefXcMwDqSNwN1dLewCcoJ/aEFxB/TxT0u5WfbL O3Dbtq1yM3ee5llzf5Z+CYIchmq6tc0/dx1pGiVjCVMD6xsTBAWlEjp9hapxoE+jmN2r01iqIFzR hpyZxgjemI4Lwx3V9OnhVrBHoek5LSqNus2IqhIcIyQTLmpr0JxDJHe6DvsjujS6qmanbWRac9E2 ZCB0hAItls+IeD9ATjYXkpPGLKG8lz7etUJ2tPTxQBTxsnGoMI7jREGeQp0qxH2Ca8oVOFQH8b7J YsgFPSiprycj1nO74lan9mSg2adVQoBGNDh0DSPpUMLoS/xIVA8Guqn4FORoIXMDwM8qED86Xhk8 y78UJoJx0/3dGOJbb/4upjl7+ZuU/rh5FmtVi+9iyhPOzmnNozBkzfXuEvHe+Wh7PWAZlIeNiQpk mgB4if6aWOg8tOnf4o7exVMSBKLjCyNPmM0gOIv/hZEY0I69xvxJSJyJ6YO30C1mQewvBgwvfnZ8 LaGTFxbCWN5RIhaaZfLI/W9FLzosLhNB0nkUodDMLiCOgE/d/3MgUPin5XvXwS5Oqe/8yJ/cbtva vHf3Vk5LmwcgAVoSswScGYzxwpKZYpM6wgz2BX0VpTEwE4wt2u0zodOXnKgp4UwSoU+ea9I/nrUP Tm7oqio0t85aNYI6iYxmU2MqUBPyMqaqWeWnGcgIYmlLBW0HCr/lDmjxdGyCJoOCgnLet/kK7A2h feOjooBQu2tIgchVqBlNOwpoIY0BUXIiDIRCWpEOl3LwTxIeKCWQ44LkSlTQqQnIfYLyD39riaUq BPzI7kt0G6Fs/6yHbmiBKwOxUwgByIvInRpFgqDlTRy5/1fD1XNiNM48KZAY13e0YAnzhblEWM0v XLL8JTpafPWLHzf4zLVYE6IsPx9Y9f7pxk/H68SNLwIH9vwGhInCXCCbXvWJ97KHCuU6/UBx0BNX IDsPy/7BVnyycVtMRJTiKfG/MKDB3MnxjcnPW9QGPkm33pZ1p/CaYVecpxNwjh+PfoH8JfZCgIVD np0i5eo8ePYosBAXVUaBw/GIP0Fk1OH6nxX0g160yJ4FgRjC6FRvCgrKLiD4d9zkcQS8Vrkt2eCZ XNEtWWB52Z7O9ucOtbtEDbU1FdUVnZKf2Jzz9/yMWn+7y4xEJAia7EhlUEFw4guIlFuTjMpUOk/P 8NKHE2d6E2Gbvx0Y8I1BtW/nsthjO5qV5IJqUodmDGcilDIzTr+mkE/Eq9zcMIT09nHgt+JJ3/Zk pG0WfGpefk3Kh+xQr34yGZSQ2IA0JcJVMibdgl79syZlRrdMgdyTiffySxxcmfv8285yX52xLIN9 JaRbAn2l+dSRTXbfKkMbufODmjd8Z8n1n1wF8lFB+ejl8i8K+6dtCt7RS+U20ec9yLYT0Bp3ureE Naxro6+8OG75qbhciUcup3t5rFsTP/8Oi/XB+/7LMLKs7JR4mDdrePF9U33jFJXHpxCoYamDMJpL ik30/PvouoUKDmDHca+b6gO8n9YHUPQjfqkJkb1k+4OL9/YvTQ/YlcjD4ikJ0lT5Tvpf+8JP/vb+ G+Ihb0QYHFv71e/4H0zVH3xX/NXnz8YVBGALEksVMzv/B6NugdgPvPPvfr98RTw+nz/rkfufOYSY avGg2D7+YQgKtTt375EICAQ2TKjcvbeLmNeKDrNgZiPXoOCOndt7jjo3cGBv6jTnqN5+sH1Hy4Hu o53lo6opLCkbh64PTwOXRCLyZOewuNK+aYmhmmDDbEr968BsxFgqzSq31qFu6Upj+YgrTaZ/CGko MRMee+o7+Zmbh3ozZazG1e6gK4/0n7nIjlPoz1TstR1EoaDhon2C0DmPoLYoiZBepJBEzQjsATCm LOUWmcwXCgEnMYXCrLLU3V665GQiYpselcilAXPt1oxfCA1KOFPLVu9ltK+nMjY1wkhCz0gcRLmK OIgQiSKUDUW2NrUreZqzFnypH7d+51dxEvk2T7QdsHwUfhPCXZom7aH7khMWBtmC1Fz5KBJlQdSS votCIA4S+8OHKJygc3GZxTqunEzUtsmB3T/z6HITxRQ2QYMefOGYbwtf3+jvyYl4SP30370Lix7O qlHVBf5uiqckEAd9d/KKppI8PpqIgz4K5jI2yWWXG4/qrpxW/sSHPLvgB0OKD7qF2vse+h4Gt0F8 PjjvWyYXyPkhVNt3uL3AVqb8DOq7TPFoLls7Vq+8X4agnIaCVBYKgnws06y/QkQEQVAQzBMKsvXL iSWUw7/oOHjwINIhOlKuTZoxtzzbFzSjSm2Aad03aclMV3zRigq2W6axAonTtKZ6ahuBtheIkIoK VxHLMnmdDjsbGe13OgXvtq2ciVAtBWnGGDQb+UXqU4sIqlkob5RxW3jpN9QjVVNN7VCaBQ20iRWG giKP6VNSLBAKwDBjEQRygE+65dolgl/SLMQCnLSrp4Bqdhgoym4lblyjmsCn4C2Wj9CZmwojVY3R uZCylM65QLkKKiP/Cf+gjVuoilJSOL/7l/LxP/cmQxfro6+k4jrRsxvrgy/+n+tsgssfTPB/4Cmr oY+FMo5ITAvnUgH8cd/NP/wHf5ENdgeD2B95YAN4i+t4E40vZB5SyJG4txdgM/0Eu57fXfpu6zy+ lyZ5dEC7yoC8YackmEUQIs4fPf6FodB38YOPKn8XGZoVJSBQTSPl3/2nt/g/kmJmFzjF3n9gkxFG pIJAxsX949ikHtyxZ2ohPzDqxBMKShCUFlS2MGYIYysve4FPbdjqULCyEhRk1QacSMgHyIGCND/U cWTTpu1rnm7m3P6C04siDuqR/sYPX8vnm2v6FpsyQ6WtCkWPtJfqMIBGU5Bqhy8jt2X1pW76kVOg BDsAV/nuTZXap5vNBr5xLhaRw7vvw2BAJp8K11U2OKqpQCkUesE60ew/8rmhqakajDWrH+19gtZ8 bcjjXjuLLm9U1n2C6Qeq0QxfCvw1qJp0mMAMWkfkrRXX1ePtoGiipn6UgpSDcl0ogqjUoVKiYud5 xced3KbbQDXKdqC2/aQFBVPZI4x3iP6THpARIUNNqC+sNYUtjhOA5d41Q5lsNOPWQfGnmMrIRpWj Ts+jV8zTuUXM+t2KGwxsLn3vF3720AN+qoegB776cyUIpEM8F/GFkLsh0GseaX4TeqDPt889xaQH sjc8vexQXCnn6+6UBkHEJBKgbskBm8uyIxc96lPN/cMfRZSwHFsmxTgN6uE/3+DgcP6st2/5/MQ4 6zRx5JtEzgyBkoAM0fCTN3xIoXyIg3rRB/468W2iB/7uNz7jU0iij+uW35tLj5r/V6rwsxjEYgpU PJ/Zl42HObUXFEzZubde8bkBDaSL+V9mbf/19tkF9hPXiGIL8zz/otQQdGJqOlPKsUyz+LIcz5u/ gJ1CLdms8tuan0Dg88c9HGFe/PMRUKR8QkMTOXi1SWZehr7qT6ah6jCADYsLanakqiYPBPXJrWks TU1qwVl8ZWygwxQN6tBXhKrERcEW2kWGrGprkBkn1YgxDaf2Vs2wSN36pNromo7ICEJ7G59N5asS HwjruQd+fN9za5b/wiAKYPMhkFZAjqDLsJBCQR0XBm/arkOrKXlOTQDU4DrASMbCgubGWzP7heCf LEhlcQoxwC3faAaH2iwUkdi4DjYTIZrVyITHUXVsDhadudPfEk8ANODoWrnQSfooSCuAkDDcQ0BW dcjmjTnd+3s8cXr8/bY3X/Q3v5n+XBwI/cU6f2/qn7VSOJQ4/UJ6yMU3KME/PZ6FI6hvO2SJQKjK KKJBwcEmN7aBBqTE3x4GCC9404fiibHiu4N69bf94q7ELFp5fk76+sGYE42lkLVIPtNJ4T3wqyMo oMWIH/BnX2SFwoGQgcr2ZXJ4gX+EBtVGIMCm5dhHAq3LIAHlbO+BglxIKepMTtp3+EAICubRoBwu qWwY1zijaaZtntmKb7rQ/EDI0MEmopjmA4+Bh+0C2mabz2HfYsXvlms/7raaWHhSf3RDLOs2F4SL aXYYHDoGdrjYNzwSxKoOAGnKW79P7Qga06RV5m8/INTuHU6B+LwHAc8G/IURsYx9O4lxgI3SQUga E/z4WIiBDJXVpwGk6iy7qtWEQgNCZZygMmKcb2ijJn4JdQwFQU3i3SA4grJgmAijgr9DOeC8rAKz yJLdWnirImsSfaqu7mz3I4vlFMzVs2WSm3vczHsf/l6geMS84nNf/XpiBokBOwRW0YjK7T0xhaH1 ENjyAIT3tD0ZyCKssFu+fb1UXvl7s24DSw2f4ESLngKZL9goBL32XfEdiWK5gJCJI7MWmGM9Tl4h lPhfGMz6T2b/efm/nBJ0lZiuLzH1Yx4WGQRSR7IgF8XzmU7sg4MhEiOGU0ejc7Hu0Z/lN5Au8C0X Ui0PEK7dt7n+0A4Az/oB+XTtq0CFDVpb3X/baCOQC628rPsm1ihJQuvezahD1Y+Un3nEQSoAhDOm zCSleyB70dAHQo2lbjWQiYD+TpsqmObWwNsQLhFLaJLZnow26qyOiZVUCIxuAvHLF5cVks1PiKE+ fWNXOoRvpvbUhShn91GhUIWFOhSkRoZFJnfaB4ouMl8hEYhSMwRCE+moKh/5Qg4/tpm0nT4EJmKh 8ijJR8LHQm7jPViJPdUQQkHbdFQCXuGiCabK7uTPq5AZxeugdCUMDf0fS4kQMpQVlgvQi3x7eRYO 51udTaZKcBNihcTNSZAhlOfWuW2dMZ4I2okdypHLT8xLBE6cHGxJyv9RH9j4OVlk2/IACP3FukAR 4b9nfixXJJ0Ce0h89QhhuBAMiF7wZN/3bhFLR406LVG0YlIPnbWzGHEwPyX+1wNkSGEQAGFQx6Yc vPo8tqb8kLDBka+k//XDbZF8DsRB9ptJ4hj/QtIPj5/ofXf+64Cxk4b23zmxVRwIUYH6Ok9fztO1 5Vgw4w4KA8MQbQdSjsno2ua23Tvdnk55xbQ4EEr5mWs7UDRPmjIfIDTrStMc+jMy2KDQrEV8s8kB mWbIrZpyWDQzTsN7HhliMShirg+0JpsaLlLfLEvd7l2UvNAVeokyfA2qviQ0HRzzqQmf2w9u5qJu zGyzBjKNrjw6JBcCqNji1tfXo0nmVmphVNMgH7ira8o7o0+WUDUKJiERZgOPDciu5Ar0IJCjN2oo F6DJhab/5JHgylm1ZLcAZX1KZdmCStXJwS2+gNLBCvC4kJuEDmlN1blcMnDVV011zrWXDWPQUzM5 9RgDoRJ/++TK3coONu185zZWLtLo5NlTec2ffAQLdWvOclM+uZ8NZCEdIuTd9d+/TWQiHT6z9b/8 r3i3O/Xtn/qjTC4dQ5YDs4CQwjb/KyEv0lMbbsmlFw0mVfjbhZnIHDu7DiBa5W9lukGq4RCSWPkP m3989fu+5bN3eCnh+8Z3NWFjddq4C4LfA1HOE7O3s9984oJLySiZ59UHgQLiub2K5DO/ivi79n+B 9ttjUgPG0S2ctwXWvO2u2YqvRgiYXe3tiIBcKyOSZbmTABRsp6l/Ez5MHNSFNti0mjdvXGFOETwS 5gn/dBgKSkcqYxkzmRkzbkZj4xRlfqByIraZn6JJTgNO34c9wzxfJhOkBSUB5knGUmJ6nz8SziR4 QQn0+574hltSchqpigvjc1uoOX5cJjYIXxL2IrRTaAjKtRpqY5LvBkrcZ0frXmCvuqoKSdEc+TVo grEMEtWAjMtfwRJHLN659xNfe064KLnNhDbfdhTcwokCy0/fNTDXEPTw5auX/PK+lRaPxtBUcOuL gMLL4mckYrI+/q1F8mewzaXUGrCVS9X2gcvwts5vYocY8c4Lr5aUmf9AufrRz70XE9BAspRGLlfb YLGmmvSxfn2SCVg6QHo7Z+5f5HFUV0MyD+RCd5N1BppQwnMp/fLgutqwLjev+y9jbCIQMvErrniH BO4hHANSEv/CYJSAEn4D65/5j1yOJbz6Vy+7gvy6A5KHAuAzn7o0kMaK57P/rmEXe5aJ7z1QPwxI 7XBVeOCR6uXN5Ud2bdKen7rdu/kp3+DClmYrNPnD94ejocGhKeUO7H0GFOQRfhHsC+bZDsz4DkZ2 pAaNFM6dOxUA8IWzxLn7DnPBbqXqS0yUWKkSAwbr0KxarEImFmh2H85MdaggV0jVRAgDDoVbgkAT +LRFWt64yEYh24YojAw7HVUAlfsbtZJZDbeWBJhQA9ziZMlfPEy4ONAe7Qv2RMJl1tsyImMCL8vk TovOk8ixEZEIGUmylyRLw7/AFsbXalLNTF0SCbVCv5rpVwMTU20uKo5M8QKuDY2DBxY9x1gi1Ois Qdd+45b7nnokHtyLte/VZ59GZoA8DuYBS9l6/OdbbiepehwOWW1ftegEEtO/86ILcmml8ogFiYt1 AITBYj0gCEE8iYIff/LmXD+MYsQU4G3dw80DopdtkomGOBBq4vIHz/8DTnwqoM1PiS+SWicoaX0s 8b8wEgdCiLzx6/feeNvtvmiYnVTZOaef+obXvfa9f3le4rZckXzmM8Letdi14q71l37hC/aThgkk weBj4liqQ30uPX5f13duv197fqXjZnAhScg3/TBU8/WffieWfkhtAxQEAseOcdr1Awd53gdylIBz h472VJW4dX/MmDH1E2azg7h73y7kQqUZyYBoXQMgUVm2HxgAFCk3xaMkJH8PzAjLxHbJesf75pRS J/rilBA9LqJRaBtvPoIK56RpFPwgO7pwObtb/A28+QsWzD3zLdNrnVcxFkZcEHBg9UMu2lEmu2H/ +pIIfal6XO/BfV17t25tgXuamu2wwrraUWW1dZlQA77kF+U7zJjRqlVbp0PN0dXjcN+kIfZHCXuE ks+C9EmD/b+ttL0uSVM2mKfEKQM/jaJNvgJlQZ8GkwKtQ9OXUmK7m4O1C80zTTSu2SFaB8uNYazP R33zk5kveuJZz5rQWIyLFYi4uqW5c33mVzXzpMlNkxsH6y034OwYZUfrPquG9XwxNA843P+qCkHA s8JfH4hICl//1R8zmxT/BfHbIA0TJfyYT2uaN+y/vUH9GPjc/OZ3fiO3BB//4qYfttElHNIotoUW DEpX3UdeMCGGpxY1JqiJLQwlYCFAyAVSI+i47NR5ixcsXrNhDc6FlACKTU2Z9FvjK52tWeXY0JNH PvturY/6MQrxWeTWTEiIa4P69MnVq01oAwIttb1Boy6000a3wjmzg5Wpi0mHtVPnU21yw0SZ0W5t JlCqA0iNa1YwlFPibwdyayagvo5Uoh5fD8q1IjbqM0LdKuwO1SgpG0U6p4wLhw2qSbnhOjp2728B /6iD76ahJoSFQCiTEKUYDF7SkG/fekL1r57ulBkqxqUkAsTdUJt/5kfBtTw3GJeaiHG2O8gjJD8J lxCG5Quq18gXvs/KBohCXNs55hk0+5Sf2LOEv/g1KtNh/6T2g54HnchF8kUxlhk0uWmDlAMpB4bE Ab4qPvGhH+PYV103RR2YTi/QiPrRxeKKysBnDoAECI0i5TcG1fgbCIUB1dKFzm10CIQkJCAEIxun LFSf+/aPYs8MsxHzLug6sGfLvm5EWCPYty+ViabUlTLU1M6Zb3LpKzM1TQl27O1xLSi1azW0JrJN hRiQmHJdyLBT6Otv5llDzVoGLCIvU1LXVV13nHShMG1vazvRBoxFFnwuI2G3V4JtAGH01qpsRtKv +npa9S8Z0VAzpx/h0hkOb/ePzRnVd9wBp+wOjlz1AULVVEJdxZ0B9rgoJi8uVjA+AXnAGzCOU1tI iSgXisuu50VRjRZCalon5UDKgSI5gBr/Z/95Z+cR4CNyB4yWSza9MLiQqYiJOCZFSayxzHm+JUi0 1LqFWKu5HZIIEWXyAKG2CVnlpQbUBUBYUT0J2kzXxzWwRH4JbGRAwU0tWzs6HRhoUw1I1racZEQz ITGDScMM02GCrEDvtFkuyA4B4XbufgYEEuWQXVYxRR0iSsIKzd2238Q0uTnqa8A8GQS9PAVKpbdU n7YXKJWyy9nU1aWsEShCz3ntkmc370Qadpzf2vLsC63oP31OKgodzEFK3rW3Ou5WqOGiF+FsZOjZ d5zQXmbgPuHeOP9kWjKzobR5twOtXEBFNeoYTfkrW7W43QpqTHlfmJxnNPgTznVt5AX0JNb3iSyk /oQZW5L6aS2EsLROyoGUAy87DqAd/ZvP/BCyif+iFVNLufarKJfxRbDt5C/rJh6Z7GVMYGMP/DvU mVlAcjnOoyC15d7A0pSlACHArD4FLUBIVXkLcCW1IVKUKGTdb5rZJCNSQrtRKAdEcxOkXFKakarU ReCcgoODrPtLXW56ak6pObqjw8WysKfqM/EVsx1I+TMbNogMk/OAPZl9Cq05eITSdfZJZ57a1Abm /WHlNig8ZdEkHr2wZbu0vms3rnJGMQc7dh90kAZz5FLpNgWjjwm0x5MbnEp2Q3PLxo0b1LP/KcOt THj8zxR93MgeOLQa9eO/qDsvsYMK6omFppQO8cO27jyxqZ5q5qrvKzyD5nKQMBeIRP4mFrLdiKDp JzXM39b8/QccQirZrI2MqrcO2CqtkHIg5cDLlANsnX7mym+tWdcKECooKBORPpDDNIFc+zKZfAMk 6Ji4I02dgZZ62LXjGbnM11a7hdGOPq+JaJvQDq31ATOVnoIDhSHbhLIfkSmN67muoS9/U90U3xAm yBcRgKKaW3AW84Vwk83GbHNMiDJRBMYyfkgXfTeAweZzYsrJ0+ZPFbYBsU8910pNPg5eNff4yy/7 WFND3QOP3/v9m37rEyCbIMKugnZYtbjZJUmE7/mzN/EIH03J65amg/59Y1GDxuDDhVdW/qUvfcTj cvWdd3Y+uaUf2y9755SpU4B9C/lavfyR2r//zq7g3eiWtq+aPXXp0la/vkw6ebTPacVzHjz97z+U NtR1lJTu4CJf1f7PTqgqmTrNtS2wCTO6b8fOVX8YuDokQfa9D+/YurZz6rSF/fkwcPO0RsqBlAMv Lw6Mrqo5erD0vt+vamlpGVUxijgx+/ezmjkIPBodbe3tR7u7R3mHechRZljIdUVlrVNFVtWU9e4p Lz08fuJxo8pHHTiw9whZJqrrgbcjR3vAv96eo5zGpe6SyBoiexzp6S3p6aUmZ0VF5tGYMbVAYE11 JQkpW/e1sne4b38HHVZUTThy5BCnWtOk49CBnft2HWjbzwhM59Chdjt7unsOHcrU7O7uBjPa2g62 7m8l8tvhw53t7W1Mtrun+0B00Ir6QCMX3MKWo908PsosORmLHjjaD7Vxq0d79+yhsp6Wl5fThIu6 hml1VeVS4ba19/D1cPhIecuBzkeeeOqJlQ8/8MiK3bvbGRSe79m7p6S3lyz3bQcPyS6moqy0qnJs aVkVZ6RnnXi4q7erq8Nxpqq7rLT++R17oHBURYXb2S0l6oWjpLOzL1g5L44TS1FY1N1bzjVP+YsR 6WCzT9RD0O/W58yUxFMyBUZvwdXUXwxwCv/PgGA3WDsdDHAGe2irsvAD8TcbmE3zSo+UAykHXpkc WPbGhfPmOc2eHZlQlni2Oee2Ku2r6aQOMofDP3taQzaGevbtOGVrI3PHcXX7tVMoFGS3z+n3vJwS XHM6ETB7UhkLUqopQ5OsQvjLHh6aQGTB1m0HkAVlLSlq7cKIpzfsJLnN7AJGJHGtJMAIu0oILI0u lEuozcwoWzmSd50Zi5ss19EZzR0XddeV/vLUbESja3erUdQQZemjK//noSee3dzcZkYrPNqw7rGH Vq1Uwg3x0xEZ/TVZ0CxCQbKD+7dRja3RKY1j4QxMYC/TLJK0c5nR97qtQTcpMURJH/0LruHPYIHQ KTkf/GO+/wA89TcCuc6jDk3sKH//8SaDDYuaj/rcz8BCpZFKj5QDKQdewRzAjeQv3nEqS6cWcpbj zIq+G7Wbu1WJUAElpNZWwwbAT1mHxKKa6oNKLqjMSqza2tbSU20TGhxyqwo6G8b08wRQE8Eq6z5C FX2iEc2AHy4E0ULv2mbd6ew1BU4FjmDPtc5AnXK29Aw5zMvepF7zdqemMSdjjRJ9JRhz/C8G+gS0 +Eurlr29Ps5RQk2mgIGSOw8fNqyqroh0yHUNYLnmaE/ZIESMc54bE+YvPmGmeVaQstFwGrWq/ysV Pf4XQ99AVWGG+kwsNL99/z3Cest/lOd/AjkfoizzrXDVD6hmTdjVw8MBX0P8/K5b4VIv+Qc7cxb1 G2kPXIxX8IPAsYeHe0OwR+jyEZ6WLPZhqhr3XJT3vR2B/4bKj33Q7TxMTh+lHEg5MEIcwLvxC5/+ +c9+7cL+IZNZCDRSIKGSUwZBHgXil2+1H63vTvYCCPkLBMrzAXlOIdbiYWV4yuaf9sNogqwje1EK ZSqiHlj3KfQ3BX0yHHiUjRtb1xV42skhRGKfmNbPV8/jo+8RkVjNKvSD0khKNhuiIDqMPCzZWTQb Tg3om79aiSv3Img7Gtp3mNRr6CX+N04olVcl8jFfBnBPcrN7WjHFj+7m+0sE/vVUHpSxTD0N/PCe Gi8eEcZPLhEHQt93kOZB7ghKggoWhtReVlCB8ngd5XXy3m+/y9w2QZlqQdQblb4UHOpzzSgtTzmQ cmAYOaBYTtahDDJZc8uqZmLi37JjPUtzXPBS/T6wad/h+z/wyFCQa4US5UKWMmeecSKrORpCZWUy 9wkGxSJUS7yu5X2hsQwFfZlPoWeC2GMizPwZbGoBKvgyX7xywOEAC3lqDidW08xu3fSzYdgyjMr6 gci3IWreF+YmM4XIU95m6gOhOkEC5ltB3pl8H/BXAmJ3z0RzlORWtksZLmW97623walGE/WixBEN xKlAOxrwDs90+rGTpwid/ikvw0ig5MxzqEJynaE60Wc6xCsDr8G8o6cPUw6kHHjFcoAwb0RbnXPi VJ0gEMsr66xyR8yY+Spc3TOAR+iT6GSp5TRVquQY5y+RVYRaYDCnAq2u92Ntu5hqM8Yi0xBTzdCR hjRB/+lHpbHVXxeIgDozwTYjSpDMMiioSlHgFZX7HutS8/IcPDBIUGFGwRhttunIlHsK1TgK9lX2 9hE1ovvrQK6fsle2RYEkLVNSKTYRbSXhYTIayIKgWiCRy9+R1yRQ7Gx/juCuPUcyFr/6amFE95qi N9Kvw8H+kAMtpdNALmkOzFWoY0HOLMSaDcRTjEjtxJQGOxROJQSmt2Jc7G0U4Baxj+yG/kkM0sAH P8f0WwfLlrR+yoGUA68wDpxzwUmzFjY9+9R2Tqam5TXSUro1mngufvIgwMbtZhEIlJVdyKToX9Eh SFMJOCdpT08RB6UyZRRZi7Doq44aIiBKz8nTFY9tFChSH90p/nOEBT7/vAs4cSF4z5+ePW/hMoO9 DDyz6EdRyughs4EnqOsfftNisohgHw791xrfa7T6fjVDTds6dX1iX6PQ2EkHiMXJ1wOnfTGAanyC iFHCQuiEyb79C6ywV4MnBpZEkgtpiEslm4jOaqmmhr9qRW+ukyz/RUvh7hPurdx595bAS+HUd5dc tKT6+cePx8fAn934hbsop2THjvqb/yNftHu5KHDSMzVjjgrVK1d1BoNeeDaxafhM6PPoiNdRt/z1 T+iZVFX6pjdV39fSGbhPxL1E4pSfNLuEtt64ud5pWp5yIOXAy54DZaVlU6dNOPjC0bXrNhw40D5u XB3ntud3Hz3SVlfXfVz9mMZx1cdNHj9vwYmzZ85ZtmT24oXHZc7j5y8+fv7kxpnTx48rqepxIV3q ahw+9R7WXhcOAEd7RnH29nRhB4JfxOGuo9tfcEFViLaKBwVeE3gLUBM/ge7urv2dR7nFTwCeEmOd ypwTGurPevPSt5/+gZMWncj5mtNOm9Q471D7c5uecxQa983p4riZk8fXV5aU1o6tPVxXU8pZgT5y 3PjaqvaGxvmTGifPbhq1cNq0stoJPT3dlOMZMXr02LG46485PGZ0+dixUyeMx2+kHs+EWbPnTG+a AalVow5091TJlwSkqasbXVpWitsDfiP0MHo0uZAiT4asS8PR7vJRZe2jKurM/QQ6JVh391ZXVI4v KaumTnXteK57e9qOdB2B81VVOI70TmkY01MyhoZS/OJZge8EXwO0pQ6nA79ZC7sPd+zct7u8Yurp Jy+cNWNCTVV1a9thfD9OXjSrekz9ifOPW7b0+OlTp/O3pKwL/5XRdVWchQOh28z70u07A1/Ab3zs ZOddVzouQLuDo0qWzXCOd5y4Bhbu5Ad6Hd23M8IbQc4QgTDX/0KB2T2rUiB82a9T6QRSDow0B8aP GTNq7qj1jzwL/jUS0WXG2KMdPWUsypPHjxs/o7yqlhMaiMPCX0KxHOro5qytKScqyu6Dz3d2oxM8 Ul1VOW/e9MMdB/D20zaVvNnauzrLeg53drV1H+2RlSkWNKDL+DG1LPqVNXgGHJ4wsXrChDE8FgrS POM5V1He1l7Rg7TYtmdT81rOX991z6rVK1ZtWIu3gtiCqASEMDSUc77nbW8YW1c/prykcnTZrElN YMbk0eVg+eQJDVOnTVk4Z3zjxMYx9bSpVvm0iQ7pJ46uKK3oxmexruaom0tNQ21t3YwxJXNmTRxd O7q84vCoUd21tYRombho7rTGieN6y6tOmDFu1uxZUxuhvVf9cM5oml5ZuZ/KxLppbKjm5Bp8nTlj 7IR6AtYcnj6lZv6seWBVSW8pQjVeiQChw7amxgVzl9TWdfMVAg/BvwMH9gFdzGv6tAlMhOmAgsiC zIgXwRTmzZkNBM6ZPXn8+NGIvk1Tx44b42LlTKqvPPGEaRS++sw3zZ21qHP8njkTZ9fXT2+oqyvc WCbBqkWpARWHLG5+krUdda8EhaTS7RZ4eKaqmbS6fsOssUxrtjChTq6BZPBSSACduBVPaixT4OtL q6UceCVxAMOZG274D1Bwem3T84e24r3H7JT5wQ7Cu9h18IhywqngRa46m5oPY+XvG47KglQKUpSB KPeoL3DVQbBN1H0yj1QSImeqk/UspBy3QguKZq0ULxQ8CN4FkczihfH31TijvmWbrbEDv0+/Ptdq QA95+lE1fxSfNq6N/pt++quM3jiaOKrgt7ztJOv8kce2nn5qkw2kC/9Wo6jDxGMQQJhoL0pQNPVr CSVsGN92VIVKkITLhFmyKNdEHCCLB0Kl502cM3uQKRAO/LtOa6QcSDmQ5QBYyEq69ejOI1tLKppK mkZNZp0dW7aY58QG27o7E5baZ5jKD/SssYWehfjBRx5hQcciRjXBP/OjWLBw9rmnnw5EzTvubNoi p1Fh7263PXnDL25B3Pzzt1+88bn/YdHHJ/35F1qoD2QCriRpOvPchSLGDnqwa+vK75OnutUodl3M O1+9bqNRPqh+IGBNyyNiFNPnb8C3u3/zAFPWvinqUOQ/wbwOIpQSNXTRzNFwSR8EeUbni0RP9bWh b4JCgZD4old8PHT4yz9V3CpWXFcvPwS/ZjaHQ19ZXlgaokToy6NxOgd0n6BJ3CUjlQgH9eNOK6cc eCVxQKlAH9uyHhTUYj3godUc/FMAa4xCN6zfzGrux5QRFjoAmHv6hectEhBySyYfy/Bz403fY92/ 8NXvAAgfemAVLhZUwKFwyesWA8zWhMKySVP3HW4PUgNRKFJHj3cDte1zcNKza7vKuVi5rWLOCW6j bnxVHbcMzYUeBXOkyXETXVqoLeseiU8f8ihccOq7+AsZ6i2opgxFPoUnHz8PBP3xL39isEQFYxpI L+8RSzdoeT+opsCkcpMguinplP3hSItot3WTHNlKCxw/CgXCuJ5wwB8BFRDsSPBLSlu/sheY2xUn 9jwsEuGyS5Z2r0nwqU+UCPE79A1cMWGNR/FOgbCQl57WSTnwSuUAWLjirvWPbR3NBEk866+zuabc 3bKWLA34HUoKtIwTEgQdBI5ywECUL+K6TZh9ovWphdtBwtpVG1bcv2DZ+XMWLWnftXXF/Xe88LxL sID/xqJ5S8obF3FtlVnoSf6eSIxhQLwC3VoP+d9dnk5EakNPy9jFF/p4YxCrngFI/gpoOQSrHKtf WNc0pXxCmUsiMa7E6ZA39+w5sGEfAIkEjBcK0QzIU0Hi30S2Kw/w3DPfkot+kdRS594dR2N7nz0R t4UCYVw8ys8vPZV+Ek+JoLIl2qU8ceOweCCkZ3CLADHBQUiaD/5Z179tq7zx2kJm0K9ObG9y0D2k DVIOpBx4WXOA9BS//sX4Dc2P+sjRWjul/pAzm+fCZqcSsGHHusee29rc1ra1YbzLLiR/eSWjt2Pe zON9gYa1nv7VrXr4k7dfUTFp7pFdmzYtv0PpjUj7B3CCjgE/fRp4JDIKOdTQr29dJXaip5PqHKRB GH+hjb+gEaQ2VDaOriypra9Rlj0OS+1n+fuwL9m7bRaP9vY8IzkVXCRZnlVA7vz2zTc/uXo17vbL zn/Lgpmn9YyfQ7c62iINqEbZvX7Ng4/dAzcY2iYLDVzv7spYD6kcUgFF4BkwRkQW9BYKhHFdIjAj 90FLPR/f7ZMIhfdeIa/B6kinGt3ScIiq0Twjgmc8LTxtk3UVy0g1qGmllVMOpBx4JXCgo6t0+aoq SzbuT0mowLGrvd2AcO/mp0iSh5kom1uZfISRFOgfM6bMRNwhGy2JhMgISH4+gFBr+sqffg0xaOF5 71R91nGg8dD2ZyQbHZm5zAYdLHMlFUlA9PFD9Cf2xqQCoKUahciUXOzYtJpsug1nvS/oTYBUtu9Z 9QkMTZmYUZk++3Sfzz6PfJw7tPNZPjgevOvnBDU9/fXvOPfUixoWLj7U2lcfIBQoAoSUr3rkZ8iF Z19wlSEl5RrRWvnATHlp6TZVKCiyDNrLuNzGVt83L6/nBOo4uYiHYqEVbQU8hR9ZA5zWwpsMNmFF 4T1bTWx/hsXTfwhDp01SDqQceOlwoKay97Wnd75/6cQ4AoEfOoUNIBZkk8uXCNGKbJLJO3g0EyDN JqW0eaDgxuZ1XLCg0xbM47RcgAInMAb8Ewr6Umku6CqEb5AqkY5D9OdqFUdBMQGC8wyETGZiGcrJ LW0t4J9OtaJQJ9V4qlOPCO5KwG4mC0a2tHQAfnbyVNfCOXEDuAXtdCKM6jx+6lidJp4CgZy9vTOQ QTlDIIzrEulaMV/8QwFlohLen05XEsRaoxBLGbSj8fJcXIvHEc3DX3s02AQU+HIU0q3VgSqQPjvf QTVNK6ccSDnwCuTAkpP3vPuMkmWTZ/tzA/8qmldwAmObn1yO3MbGFS7pDeMaCaqC2wOWMmwNurOk S6ea41y4bZtDQTItEJwaIZK2lANR4GiwK8aKrxLJYYZbBsP+xYCspxMfxgC2QZ0SK9kaHH90E3ab GPWIquCQUKhDTZALdYJwPPUruA4rG3nETKccf+pr3nvVktPfg5znn42NNbrlQpLf/NlnwRZ/BxQn hXXbD/jno80HdK59YSwoSKsHt3ahjA1Vo3L48+eAnAcQBoWvv8TyS/T7GcRtREkBQdJ5KhF9+/HH V+V6K2hZyREIZHr4qroJdqex0Z36tHBsS5xRLsKgP4mqAX9daYWUAykHXvkckJp0eXO5pECJgMDY 7t1OoEGaAQUR6bhlV085KywZvc8dBV2z6NIEA5N1DB3S27x3fdqXPhkLrAW9gi2xwbJbu2UgB6MM uStQDQUmQ5/alDE/Wde5KL5RF+wXli8unZG16dk2uQSrRovKidCmay6AHjp/wxvPra9caBXyTPPp bc2tT//2Ty68LLEOgqO2FeNHCISDZGVr//r1g2yeq7rfbZ4+rdpwjZuf/GCywzTXtJuUAykHXuYc 2Phc3Yo1jWjzgKi9j92NkScBM0mJ0NE5JpMkNkpj68I9d7bKTLQE7Wj/nUKkRsvuhID4qvPejHyz dsXvsIiZcOobBISmdJUpTYCOxkWrTEmgv5X4SKF0oQAhSNOx8aGaeefUn/CaAt+DDGF0AKNAqWwy sUDBLwIHiQPdc3J1BbyBgvGniRb+CHNwddboRtvty0Ph7NMqNz/a9Yd7b0J8XHzBOK41VmLP9CMy hMcBEBbIh7RayoGUAykHUg70ccBEw/W//wUyHG5tuLgROE0Z25WrzwFhFCktPCJQRCnKU5LxKif7 a897PV5xcpwACGliEqdsZIRnA24N+qCocYWCcplAdEOEAghRafa+6ooC36ipMdnS8/0QAEJt+8nm BfErsNi0/tWD0FQ2LzJ4iRNgFjH2KFGqsx4wmWFeJy89P94V/ecC1NLe3sFZshTIqbRayoGUAykH Ug6kHHhZcOD/A2uTR9mtyQXnAAAAAElFTkSuQmCC ------=_NextPart_000_0000_01C0CE5D.B524C4A0 Content-Type: image/jpeg; name="image002.jpg" Content-Transfer-Encoding: base64 Content-ID: /9j/4AAQSkZJRgABAQEAYABgAAD//gAcU29mdHdhcmU6IE1pY3Jvc29mdCBPZmZpY2X/2wBDAAoH BwgHBgoICAgLCgoLDhgQDg0NDh0VFhEYIx8lJCIfIiEmKzcvJik0KSEiMEExNDk7Pj4+JS5ESUM8 SDc9Pjv/2wBDAQoLCw4NDhwQEBw7KCIoOzs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7Ozs7 Ozs7Ozs7Ozs7Ozs7Ozs7Ozv/wAARCABXAqMDASIAAhEBAxEB/8QAHAABAAIDAQEBAAAAAAAAAAAA AAQGAwUHAgEI/8QASBAAAQMDAgQCBgcGAgkDBQAAAQIDBAAFEQYSITFBURNhBxQicYGRFRcjMkKx 0RZSU2KhwVWSJDM0NlRjgqLSNXOTQ2RyhOH/xAAaAQEAAwEBAQAAAAAAAAAAAAAAAQIDBAUG/8QA MhEAAgECBQMBBgYCAwAAAAAAAAECAxEEEhMhMSJBUWEFFHGBsdEjMlKhwfCR4RUzQv/dAAQAKP/a AAwDAQACEQMRAD8A7DSlK8wuKUrVX69t2aKkhPiSHiUst9z3PkK0p05VZqEeWQ3ZXMt2vUSzshch RU4vg2ygZW4fIVUZOrrnOUvwSiGyDgBHtLB/mVyHurSvvLlOLnyV+K6+MLWpRBWgE8P5Ggc8Bjca 9PONxA05MeLWcGOjwwXT5Nt8mx54z519Lh8BSpLdXl/eEc0qjZJStb7yVS1uyQTxLqlKBH5V6Q1C 8fBYbIByU8M4rU3O73C3OIS1ZWEvOJ3J9bPjvgDjkhWcVphq3UKSlRDRScKAMVGCM+7lXfDCzqK6 aSt5+xk5qJa0uSYrpVFfejjdgFC1JA/tW1h6wuUN9DcxCJbSuiRtdA8uivhVatt3mTfEU7Z0IcZO HDAUWX0+ZQOBHwrO2Y0llTsZxt+Jzd+yAAP/ADWuQ/8AySAaxqYeMumrFP8Avkupd0dNt1ziXWKm TDeDiDzHIpPYjoal1yyHLehz/XorymnxhCitWQsdEOH8QP4V8/OuiWa7M3m3plNpU2sEodaV95tY 4FJr57GYF0OqO8fodEJ5tmT6UpXnGgpSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUp QClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKVFl3GPDH2i8q/ dHOtJKv0h72WR4SfLnXTRwlWrulsUlUjEsD0lmOnc64lI8zWskahZQSGUFZ7ngKry1rcVuWoqUep Oa816lP2dTj+d3MHWb4Nk9fZjv3VBsfyioK3HX1jxFqWfM5rHWRlAW4AeXWu6FKnTXSrGbk3ySfU CWgQfa6g1GWyW8hXMVsi5tAS2M+Vai8y0WiAZTxyon2Ud6jPbdlsqfBjkS48RIVIdS2FHAyedeTd YcZaPElNpzxHGuf3O6ybg+XnFYOcJT0AqKuSG2gSkL7cKwliHfZF1SOxtP8Ajth5pxK0qHApOa+B xaXeABPLnXIReHWG8MvuNA/hSojBrGb1JLiVJkr3cyrearrrwTpvydtZkuIA3ZwOFfX5W7gTy6Cu ZWnWspvKZTpfRywo8RVnh6lgTU4DwbWeAQvnV4yhJ3DckrG+RIyrCjmsqSVK4Y21rkFJO5pYUMd+ dbGDzClHA5gVpKyVysW27H//0OlMRQkBSjnPEg1JACRgDAryVJQnirlWFySEuYHIc67OqTJ2iSK+ FQSMk4FYUy0E4x/WsilJW0SCMEdajK1yLngykgkAE+dKh+Ij99PzpWuSJTMywUpSvnDU+KUEpKic ADJrmlxnm63OTMUpSkrGxpvoGgrAH/WoKJ8kirzqKQuNYpK2xlak7AM458KoYaShBjtZSpIIT5Yw hP8AXcfjXu+yqaUXUffYwqvsR1LEdsSSwl59x3wozI4h94DBUR+4jkByzk18uEe6aElN3yWW7g/N bKFqdGS05zGPKt1py3ol6ydJSVMWWMhhsEcPEVxUr386kXdpF/8ASJDtjyfEjQWFPuNqGUqJ4DI+ Nek8QlV02rxUW5fDsvp8zLLdX79ivuRcwYsSNKEu66gc/wBJkpOVIa5qA7CtkizRpetbjZC2lEb6 LS0hI5J5YI+PGorOlkSNfTm7NKXbBCaCklr2gFngeB6YPKsUaz6iVr2VDGoAmWmKkrlBge0jP3cd PfV88Gnlnbpvve6va3G2ysiLPuu5GUALKq5OSUxb5Y3PCKt3F9I5BX73Cs0G13LVZl6qgqTBkJOG GGgAHCOYV3z51MtemY0P0jKj3NSrktxnxkOPD8XMkjka3GnU/RGt7vaEIKI72JDSRyBPPFUq4pRj J0t5Wzelm90iVF334Kqw80+w28yyG8ktrjEcG3PxNcfwnmB0zwrbaeuTlsvDTjilFmTtacKjzGPs 1kd8eyT/AC16v1vTE1fIjNja1c2PHQeiHU9R5nFQ5CPHSACnOCRwxjcAsYx2JVV3KFan6SX1/wBk JOL+B1ClRrc+ZVtjPnm40lR9+ONSa+PlFxk4vsdqd0KUpVQKUpQClKUApQnAya0dz1haLWstrfLr g/C0M499Ck6kKavN2N5Sqan0k28uAKhvJRnircD/AErZxdYW6dc2oMNLjynRncBwHvoYxxdCbtGR v6VjeeajtKdecS2hIyVKOAKrc/X9oiK2s75Kv5OA+dDSpWp0lebsWilU+P6R7a46EvRnWkn8Wd1W uPJZlR0SGHAtpY3JUORFCKVelV/I7mWlVaXr+1xZTjHhuueGopKk8jip9i1RDvzzrUZtxBaAJ39c 0IjiaMpZIy3N1SqxP15a4E12KUOuKaO0qSOGaRtd26S086GXktsJ3LUf6D30K+90L5cyuWelVD6x 7X/Af+VPrHtf8B/5VBX33D/rRb6VXIGtbfPS+tLTjbbDZWtauQHaof1j2r+A/wDKpLPF0Ek3Jblv pVQ+se1/wH/lUy362gXFboQ06hLSCta1cgBUCOLoSdlJFjpVQ+se1/wH/lT6x7X/AAH/AJUK++4f 9aLfStVZNRQb8hZiKUFt43IWMEVEvGsbdZpnqjqVuOAZVs6VJq69JQzuWxYKVUPrHtf8B/5U+se1 /wAB/wCVQZe+4f8AWi30quT9bQLcWUvNO73WwsoHNOehrzbtc2+5T2obTLqVuq2gkcKkv71RzZc2 5ZaVpb5qmBYnW2pAUtxYztRzAqDC13AnzGojEd4uOq2pzQmWJoxlkcty0UqNOuMO2s+LMfQ0n+Y8 6rUr0i2xl0oZYdfA/F90UJqYilS/PKxbqVVWPSFaHI7jjiXG1oGQ2Rndx6Gtj+0sNNi+l3UrbZJw lKvvK91CI4mjLiXqbmlVD6x7X/Af+VPrHtf8B/5VBn77h/1ot9K0lm1RDvKXltNrabZGVuOcBUCb 6QbRGdLbIck4PFSeCfgaku8TRUVJy2ZaqVXoesrfcLg1DiNuurdAOQOCff7qsNDWFSFRXg7ilKUL ilKxSJDcZlTrhwAPnUpNuyB6ddQygrcUEpHMmtDPvq3MtxfYT+91NQZ9xdnOZUdqBySKh17mGwEY dVTdnNOq3sj6pRUoqUSSepr6E5FeayJICedek9jE8Y44r0EedeuBFAEg8Ki5Nj5sFS4rSCgnhu6V G4Cq3d9WerOqjwWvEcScFxX3RWc5qK3ZaKuy4yJcW2RzJkuJQlAySTzrkeo79Kv81bqisMg4bA5A V4uN3lz3D62+p3jwT+Ee4VGaRIcTuVhKBXFOeZm8VYghMlxW1SlBPc1KXHCI6Wi8SoccDpWRas+Y HKsLm4r3oPvzWZY//9HnzqSlRG8qB5VjQCtW0Ak9q2rJQfbKElQ61lCGkBRShKcjhgcTWwNMQ4gj gR2rIh94HgpWf6itmkJCUpKdyunlRaG0HCQN5+8aAiN3m4NZDcpxIPZRFba1ar1BD9hiavbz9sBW PnUNllkq3BIwOuK8uLGTs4DlS7BOc1pqFuQ64Zznt8x0+FQndUXl5RUu4PZPZWKxKIdThVfDDS5h IwOpNLvyLI9Iv11bWHBNkAjkd5r25qi8uhQXcZBCuY3kUfaS2wlCRuTjrUZEZC0EhPIccUuweDcp hOfWn/8A5DSsXgdqUB+oqUpXjFzTaqBNkVjo4kn51T/EQJac8duN3wdV+oq+XmL67aZLAGVKQSn3 jiK50iR48fJ+zUoZI7Z4Hj5LQfnX0Ps1qVFx8P6mFTkyWy13y4ahvgtN6FvCJAK0eEF7sjgeNYIt n1H+2kuEnUQbnerhanywPtE55Yqdarn9G6liXRwBMe5t+rSenhvo4ZPvH51M1jLRpzVFs1FwU0pK mH0p4qUkjIxXdq1VV0klvHbZcrs/O6MrLLc01rsupHdT3SMxqHwZTYSXXvAB8Qe7pXhiy6jXrqTE RqDbORFSpUnwR7Sf3cUM2/y9ZNyY22zfS7RQ2t0eJwHHPvPCvca06j/byVHa1ADPRFStcgsDC05+ 6R099b5prM5Siunx8E72XFyuzta/IcsmpP21ZiHUO6b6uVCR4A9lPbFfRZdSL1oqInUQM1EfcqR4 I4J7Yrxa9QTYGr5lzvbXrLTA9Vckxkew2RwzirHo0i53W7amWNrUhfhsFR/+mnrWNerWoxcpJWUV wlu32/0Wioye3k0sm33e3autabvdxcVbVLADQTsT34V6aUkOoJGBsb49vZJ/Iisc65m632Zd2gdn +xQSev76/dWN5xx0JbYT7Tpw2c/vHagf5Eg/GrxjNqOfZ2V+3l/yE1vY6Fp8FNhhhXPwwa2NYorA jRGY45NISj5DFZa+QrSzVJSXds61wKUpWZIpSlAKHgMmlaTV10NqsLzjagHXPYR349aFKk1CDk+x VdZ6xU6ty3290oabJDryTgqPUDyqn27S2pNWJLluCYMHOPWXubnmkdq2WmbM1fr03Fkn7EfaOD98 DpXXkhiHHCUhDLLScAcgkCpi0tzysJT95br1d/C7I4Jf9J3nQjkebcLimZBfUW1qSCClWMjh8KuX o9ajohyNUyFFMZpBQznhuPU/kB7zW8lagtGorymyyLYJ8UKylfMbxnjjtx51qteS2ISIthgIQywy nxFtoGAM/dH51aUk1fuWqvDxviIbuO3pc0GptVTLu8Sd/glW1mM3+I9PeawRvRprG7sesPzGLclX FDOMqHv7VbvR5YI7m+8yAl1xKvDYSePh9z7/AP8AtXO6XWLaIa5MpwJSkcE9VHsKiMspOFoxcNet u3vv2OFMQLnanpFuupCn4y9u8cQod66npRMxWhViO6lt1QcDK1jIScnia57Olyb3eiW0BcqY7hts csn+wq+aomp05piPZ468uraDZVyO0DBPDqah85jmwzjGdTE2tHscuSDbgiO7JMpxbhT43Ldx51dt AOOC4TG2iErXHO1SuQI71rbpYxB0LGujycPSJjShkfdbJ4fPn8agW+4u21by2SQp1pTec8s9aS8n NJPD1YVJ8vdmKHZLjNujtsamJlS33lqL+32U9zjsP71s7noG5WHSqkSL0krckhb6wk5WAPZQnjw5 qJ+FW70d2T1aEu7Pow7J9lrP4Wx/cnPwArR69vnr9z9SaV9hFyCRyUrr8v1qczR2vLQw7qSXVL+S rsWp+7n1GNLbiKUnJfdPBIH96i3zStw0c1GnTbobih9RCGWUnjw5nyqRb7TqefeWI6bMtqI64B47 g4BHU8+1dyEKOWWm3GGnPDSEjcgHFWV4bM0weFapuFWPP+TkzljmD0eqmqkeouT3EKUhYyrwwchI 7EnGa1DFsfuyvUmJbcRS0n7dzknhVo19eBLuYgtKAYic8ct3X5VTFQtVOScMWJwMKV7LrnAY7njw qFeT27HLUi6tZRpRvGB9vOlZmk4rFzm3sTmXF7UNMpOVn58qsbFimSPR8/c23/UVTdpysEq8IcgB nmePwxV/sLlknW9iA0Y0xcVoBWWeAPU8arXpBu6FOtWiMUhtgAuBPIHoPgKOffudlbRp09dJXtZW 4uUyLAdnLRBYkIZWtO0PO8kjuaiixSrJPWh28t3Bso/AOAPxr4sXVm4NiRBUzCeSSy8of6zAHLjy 41ubNpGRq2YGkzTFiMYVI2j21g8gO3I8ahXXS+55lOE9sO0k5b3Ld6MY26LMuA4oWsNIPfHE/mKp d6YkN6juLsmYmSt13PsDCUDsK6hdHYekdK+BCQGkto8JhA5lR6+/ma5DKTdZIUq2QHJz+7KwkfdB 6moW7sjqxUMsIYSluya16P7rfFIkR9QR4xdTkR8ElIHfjUXSVvN3v6bClbshLSj48kggFI547Z5C rJ6P3XLRJkS9TNNwFrRsYaWklSu58u3xq/3OVAsFpfuLEZhtZR7GxsJKyeQOPOruVlZnVCjSdK1V JZebfycn1THftuobjJmSxIOchtsYS2non31n0y6BqC3OgkAvIPwzWxtFocvNvvN2lguBMd0IJ/Gs pOT8Pzqu2iUY3qkrbuLe1eAcZqj3Vzy66d44iW2Z/sbfWkWX+2kt96Wl7xAlLTDY4NjGAD3JGD8a l2zTEzRktWobzOQ+4Uf6PERkBKz348cVM0bb3b9qJ27zBvQwsrOeRcPEfKoOtLx9K3x1DaiWIxLS R5jgr+uanM7HVKooU5Ylrqk9vQ091ul2vtxShttcya+fsmU/dSP7CpB9EurZkcvv3lhh7GQygHA8 s10LQ9gjWy1JnEJclyxuW5zwnokeX9ya2N/1FEsUUrcUHHz9xoHiff2FFLLwddChTpU9Wru3u2zh 1mdfmXFqxLJXci4WlbU5AI5n4DjXQPSDanYlktEJiYlmNFWCpvGVuqA79uJqzachWeWgXyPZ24kl zO5wjie5B7VQ9X3g3e9rLeVtMnw2kj8Xu8yaOSv0mFdUsPRbp7ufBp2rHJvwXGj3BqBtG4uudePI Vr7zYpWi34zk+6KuJkpJbbZSeA7ms0OBqZy4NJl2gw4ZWPFfeHBCOpPHjw6V2L6T049bVzSmO+zF ASVFkHHYDIq18m0icPQUabp1kl89znUtblrszNtBU2/JAflIzxTkeykj3YzVal2eWIjt6RdQosOD MFKDuKe4PX5Vtdbxrmu7yNV2lDky2SMB1JTxbIGOXbhzrHanFyX4i2cpU8UlIPPj0qPy7rhnNVTo VVJLNB7L4Ft9FkBycHL4/HcYSnLTKV8Ce5/t8DXSq8toS22lCUpSAOSRgV6qrt2PcpUoUoZILYUp SoNQTgZNVW7z1S5BQlX2SDgDvW9u8gx7c4ofeV7I+NVKvX9nUU71H8jCtLsfKV9r6EKJxivZOY81 9AKjgDJrBLmxIRIefTuA+6nia0MzVElsYjJSzuPsqPFRH9qynVjEuoNlifdbjJ3PrDY860E3V7Md xTcZnxCPxE1WbrfHH39jrqnHDzyc1q5E4tuhGz2jxKjXJPEt7LY1jSS5P//S01w1BPnK4ultHZHC tfuUG/FUcdE+dfWHW3W9xGdnyNeFuFw8Rw6DpW7be7ISsYilCxwBCs8zXvLhzk9MCvTccrXknakf 1rK6knakn7PuKgkjIAUoZzwr45hB2q457CsyiMbQnA6GsaUAugnHDuaA+pYSjG4EE8efKiVKO4pw Bj7xr4tZU4TgnoK+pbQpIBVjHMZoD00AGyUoyo81E1jUpLacYBUTxNZ1hTbZSCgg9M1HLOE5PWgM alhQ3BR4dKEAK2knCqyJGUlRAynkK9Ijl7BwdwP9KAwlK9xTnl/WpDSeG1XPr50LSW1EqJODwFSm SFoyBjHMCgIw3LPh7AD1xXwp8FJAAyede0I9txZWeeMUWg89u4e+gIZbBPWlTAwpQyMcaUB3i/3K Ra7YqTGjGQvcE7R0z1NaFWp7qFEBuMR32OfpW11cGxYXHHpvqjbakqUvBIODyOOPGqzZ2rfqB9xu 3XNDq2wFKSA4AAffVsHSovD6k43ty7MpOTUrI2H7UXb+DG/yOf8AjVfdjOIRNlKwtTjm9EVplZyF ABYycc8BQ8x51tbxAh2FttdyujbAcOE53kn5VBgybLcZiIka9trdc+6Nrgye3GvRoKnCLnSi7eUm Zybez/gjLhqjtuxJ8d0xX0JWvPBZH4XU/wA6eSh2waiIjxlXLw9WSnZcFTPhQZiOLaexOOShVwla cTBZMyS/Fwx7SVvKVtQe/P8ApVXXdrM1KcUm9Q20uH7ZhMZa2XO5weR8xW1GvrXyXfql9ObMo1bk xz5DyLY3b5slsyYCg/a5oOUPJTx2k9DivDmr47F8mXplYL0m3JbbR1DvAcfdz59Km/Q2h9RL8GBd 3Y6gN3hcdnAHO0H9ar5sOjkPHdqp0pBxt9TVn55ropaMrxmpX7rK+/w82+pV5uV9TaR1LfsybDDf QmKpXiXS4OKwjJ4lKSedYmHJQanW20TVM6fcX/tLyTuweaUdVZrNDb0FBjlP0s7KcAJQl5Cw1u6E pA/vUtNxt8F5qVLvMGQtxG5gllwIbT/KkJ4H31Vys2oxfzi935+P0XZi3l/uYw0VR1JZiutxYaNh aQNxbR+7w5uK5ntnFSIIchSoswLbdcbSVqbcacSEOHtgHglICR7qis3PTrI2N3SM0gkkhHjcSeZ+ 7xNTUy7EqCZwvCfAS54RXtc+9jPLHnVJ5krOLs/Rllbybn9qrn+5F/yO/wDjXlzVlzQ2pYajKKQS AEO8f+2tF9L6a/x1H+R39KzMLtV0PgW69xlPLylIcW42c+WRXK8LRjvKnZfB/Yvnfkv9slOzbbHk vNFpx1AUpB/CalVEtcV2FbI8Z57xnG0AKc/eNS6+Yq21JZeLnSuBSlKzJFc49MTsqJAtk1ttS4zL yg/t6AgYro9YpMZiYwpiS0h1pYwpCxkGpi7PcrOCnFxfDPz/AANVRIrqZUWf4Do5HByKkzdczruU xW5Um4uq4JaaTjJ88AV1F30Y6Odd8Q2VoE8wlRAP9a29q01ZbIMW22sR/NKePzNW/D9Tzo+zYR2z O3i5UtAaMnwXxe72styVJIYiIV7LSTzKj1P5VUPSHMds2upJnJc9WlNpWy4BkDHAiu4VBulltl7j +Bc4TUpvoFjl8aKSvutjrnhqcqWlbY4lbtci1tlMK7eEhXEp25H5ViVf7rqiYGLezKuclRwCRhCf MnkK6o36MNGtu7/oVpXkVKx+dWKBbINsZDUGI1HQOjacU6FwcsfZ0ErSk2vF9jmZtrfo4sX0vc5a F3mWfDQ5jKGe4SPd1qrzNXx7xNQ7Knh904SkbT8sAV2q86Zs2ofD+loDcvwvubyfZ+RqBG9Hukoc luTHscdDrSgpCvaOCPeaXi+TWtgo1bK7SXZcGp1+hxr0coVKOXGlNLcJGMEc+VUXTTTGpLy1BiPB YzudKQfZT1NdouVsh3iC5CnsJfjufeQrkahWTSlj06t1y029uMt0ALUnJJA99RdW3JrYSFWUZS7F K1pr1uw3UWaFMRFTGaSlQCcniOA+WKog1Db3JAX64lbilhX3Scqz7u9dmmaA0pPluS5dlYefdVuW tRVlR+deGPR1pCM+h9mxRkuNqCkqyrgRy61PR6mdbAxrSzSk/sbiz+MbTGVIUVOKbBUSMc/KqZr/ AF4mwXFq3MTEsOhG9w4ycHkK6ABgYFaC46G0zdpq5twtDMiQ595xZVk/1qqt3OudPNDIm169ziqt R21bpdXMQtROTuSTn38K20n0luy4q4zt3HhLTtICCMjtyrpf1aaM/wAAjfNX60Ho10aCCLBGyPNX 61b8P1OGPs6MU1GbV/U5+nU50lbmUJlJjyJw8f2k5IRyT7s8+PetA9qSBNlLdcmBx15e4+yeJJ91 douGhtMXWUZU6zsPvEAFairkOAHA1hY9HWkIz6HmrFGS42dyVZVwPzp0epM/Z0JJLM7Ip/pBkot9 k00qc4UEtuZ3JxtJSg7cDty+FaXS2rWYMqQ5Em4aQ34j+E/hHLn5muuXnTln1ChpF2gNy0sklsLz 7JPPkfIVAa0BpRiO9HaskdLT+PESCr2sZx186dNtzWeCjKoql3dHJ7rrxq8lHrlyStKCSlISQBn4 V8tmvGrOhaYNxQ14hBWfDJJxy6V1P6tNGf4BG+av1p9WmjP8AjfNX60/D9TL/jlmz53fzc5zBvg1 DdF3J+Z46ISPHfcIwEpTxA4jqcD41Fuuv27wkNyrklTSVbkoCCAP6V1pjQ+mY0GRBZs7CI8kgvNg qwvHEZ41F+rTRn+ARvmr9adHqS/Z8XHLme/Pr8SNohCpeiVqBy0+F+GNmBjv5551yS0XCNIZiQ2n QuStKUBtIJJV2r9DxorEOK3FjtJaZbSEoQnkkdq0du0Fpi1T250K0tNSGzlCwSSD86hZbNM0q4KF SnGDbsiXY7MLLYm4TePF2EuKH4lnmfnXCHbuIFzmQLtuZksvuBSlA4V7ROa/Rtaa8aRsF/c8W52x l9wDG8jCvmKJriRpXw0K1NQfCOQR/SC7b4YYjXlSGk8kJTn8xXq0Qb9rqd/orTrUVR+1nSM8uyR1 NdPh+jjSUF0OtWZkrByFLJVj+tWRplphsNstpbQOSUjAFTeC4MYYCnG2ZuVvL2OTah1hH0y6NNQL j4bMNrwnCrKlKJ58fjVUY1Nbo0hEhuYjxG1BSSpJOCOR5V2OT6PdJTJLkmRZGHHXFblrUVZJ+dY/ q00Z/gEb5q/WnR3uRVwEak87k/t8DmU30huXZj1N+6eIhZHsJQfaPTpUvUV3ZsFojWN93D4R6y+2 BxyeITw7eddGj+jzSMSS3IYsUdDrSgtCsqOCOIPOpbukdPvXNy5u2tlyW4CFurySQRg8M45U6PUe 4Rd80m7nHYmvRDta4jFxCWHActFGTx5jlVl9HNjlXWei8yYy40GOcx0rGFOq6Kx0Aq4RPR1pOFK9 ZZszHiZyN2SB8KsiEJQgIQkJSBgADAFG4rgtSwMKclJtu3F+x6pSlUO4UpSgNRqM4gtnp4nH5Gqf KusOI0pbjyfZ6A8atWtIb83S8xMZRDzaPETjrjiR8q4SuW+vBeUlSeozxFevg6+WllRjOneVz//T 9q1lHCSpEdR44BJqBN1dJltFtgeCD25/Oqw7MbQrCPbQfmK+Imtp9otnPSup1ZvllFCKNkmS4+6f EWdx5En86wSJjEUqC3g4/wA89B5VGTNKyQQDw5pFQZTrIBaDa9+ck1hJmiIzriluqcCsEnPOvKlr XxWonzNeTUqI+0he91oO4GAknAHnVATrXu9XVvB2k8AK2TbKTnKQM8hmtTDU5uWG2gr2vZyeVbFU 5iO3sdbT43UDiB76upCxlcZU2MZCjnlmvqWiniXMDseVQm5MV7ctx7aR0OeJ8q+bl8ClSVg8dijy qxBKdShQ2+IgAedY0NjbuUtKgDw486hkK8XCilIJ59BUl2NFKfspQWrhnhgClwZVICgCkcTyAVyr 4W22QPEWgKUeR61jDbaWnFeM2lWfZSATUdC1BYU4yhxIPU8akElxTY4LUlR8ulfUICm8jxXAD0Sa xTVIeKVNtpScZyk9O1SU3C5swk+EsNsjkTjNNu4I5fbZBJbKSe5qRGdZcIXvKcc68m3OP2oznZDW wKAA3Ddn3Vi8OHHZKjOKlpIIShPA1VuxNiS85GkYwFIUe/KoaJDkdRTuQeOOBrXSJjshZPHAPDA6 VITBJgiYFgDJAGck1CYJhc8VZ3gJ3cdyTwNYzJYCCElRKeABPA1DKw23gpUc9DXktMEpw7gnnkcB ViCZ9KqTwDYwPOlRD4TZKQtsgdcGlAd49I3+5cz/AKfzqmeh/wD9Xn/+yn8zVz9I3+5cz/p/OqL6 LZKYcm6yVnCWowUf616GBTl7KqpefsYVP+1GD0n3Q3DVHqaCSiKkIA/mPE/nWgfhytM3qKXSA4gN vDHQEZqOqY9MvS5wZU+4t0u7ME545xU7U95mX6U3MlwPVlNoDWQkgHHv99fRUaToxhRS6bb/AB/t zlbUry7l+9Jdwcl6LgPsqIZkuJUvz9k4H51RNOxNMvsuqvtwejuAgIS22SCO+RVjgaliI9HCGLhA M5DUnwVJKsbcjKTn4H5VrLRE0Xc0uqmSJFtUk+yhSioEe8Vw4ZPD0JUmmkm947vn+9jSXVJM8QdK syrun6KvsFSA4C2FrUhwj3Ec/LNbLWWgYenbIq4My3XXC4kbVAY41TrgmLHuribU+46whf2ThG1R 866r6RitWgWS5ncVN5zz6VOJqVqdehae0nZq2/zEYxcZbcFL0No+NqoSvWJK2fBxjYnOc1j19aW7 JcYlvacU4hmOAFKGCasvod5XH/prU+ln/edv/wBgVSGIqy9qOk30pcfIlxSo37mO1WPRb9rju3C+ LYkrQC42Ek7T8BXvUsC0W/R7X0JMVMjPS1FTikkYUABjjXi12/QjtrjuT50hEpSB4qUkgBXXpWTV AsrWj2I9ikLejolKK9/MEgVe8tePVJ9XDW30/wAEf+XwedEaGiaotj8p+U4ypp4tgJTnIwD/AHrS aqsR0vfjDakFzCEuIWOBGc/pVn9Her7Pp60SI1xfU244/vSA2pWRtA6Dyqva4vsXUGolzIYV4KW0 tpKhjdjJzj41rRlinjpxnfTtt47ESUNNNcnadNyXJenYMh05WtkZPetnWo0qhTel7chYKVBkZBrb 18HiElWnby/qehHhClKViSKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKUpQClKUApSlAKU pQClKUApSlAKUpQClKUApSlAKUpQH//U7DSlK8wuKUpQHwgKSUkZBGCK4lrnRi7NdXJLLobhvqKm +BO09jXbqhXa0xb1b3IUxG5twY8we4rSnUcGD85KhqJ2trJcSPaBHP3V4RFlDiI7yxn9w10+5aSj WibvdjvONEYQ43gfOsPqNsVgJYmDHQOCvShmnG6RRuK5Zzlx11hOFx1ozwBUnFQnypxw8ck81HrX QL43aIUbx1Q5C1A/jWDVBuEht54lhjwEHiE5yfjVWmnuibprZnlploMuOuLwU42J/eNG/BUMuApI HAjvU/TsSJJlFU9JLCeeFYNXafbdPJtWY1td3uEJSrcOZNSoyfCF15KS0+ttjw4hC3XCMkDj7hWR 2wXdthch+E8lscVLIrZXizMw7tFbixnm2iBuVz41a41uiOwQzO9aeJHE7uYqYwleyRGaPdnMShO8 Ecu1ZlKOwDCgR5V06NZtPRjxtbqh1yc1KciWBTgX9HOqOMKq2SouxGeD7nJkl1R/1KlAc+Br0Mlv CUq58sV1dDNjG9KbW4E9MKFeVRdPYx9FK4c/aFTkqeBmh5OUbXCkkIXnnyryFOFB5kjr2rrSYunE IJTbFhZ4EbuGKjSW9OxEDFqxvUARuBo4zSu0M0Xtc5m0U81JUrPwrG84or2pKggchmuxtWuwyMbb chHDmpYFZUaWsLqiEwmFHPH7UVDjNcolOL4ZxdDZKCVLIHYUdZAZKkjOOPPjXYpWlbQp9CIzcNCS PtMuZ5dBVc1TaotijlxhiItb48NAbWSoE+WKo83gtt5OdALzvAI2962jaTIipQhOxZOdoNby0aVu rSjKfjMu70/cdVW1TbZkWUC5Cith8+zjjg9qs6dR9iuaPkpTsRzwg4lClJPJWOBqIuHITkmO5w5k JOK6cLdNU3hTDGwdA3wFZmbPMcBbbcUEq5jw8D5mradS26IU4Pucmwr+GflSuv8A7Hg8TGQT33p/ WlVyz/ti2xeb7ZW77anbe+tTaHMZUnnwquQvRnBgRZkdma/tloCFnAyBSlRSq1aMMkJNIq4pu7RJ 076Pbdp24Ga04t9zbtHiAYFbW/6bh6htphSU7E7gpK0AZSaUpKrVnUVWU3mXcKKStY0lu9Glsgx5 UZb7z7MlICkK4AEcj7xk1APoftWeE2QPgKUrdYzEptqo9/h9iunDwTLZ6LLLb5bclxTsktq3BKzg Z8xW+1Dp1jUVrNvkLU0jcFAo5jFKVjOtWnUVSU22uCVGKVkiLpjR0XSwf9Wecd8bGd/TFRNR+j+H qS5euyZLza9oTtSBjhSlFWqqq6ym8z7jLG1rbGq+p+1f8dI+QqSPRbbk21UETH9infEJwOeMUpWz xuKfNR/t9iNOHgjfU/av+OkfIVJt/opssKWiQ647J2HIQvgM+dKUeNxUlZ1GNOHguqW9iQlKQEgY AHSvu09qUrztCJpcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPal KaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptP alKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2p tPalKaERc//V7FtPam09qUrDQiTcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalKaERcbT2ptPalK aERcbT2ptPalKaERc8PRkSGi26gKSrmDVUvGkZWxa7WQVHkgkDHxNKVtSvS/KykoqXJSJmhdaTEL S9Bbc3//AHDeB/WtO56KtYqWT9GNn/8AZb/8qUq7k3yW4PTXov1m2kpFuQASD/tLfH/uq0W7RepU tRW5cIAIWVuHx0HjyHWlKmM3Ehq6Nq/pS7yXEgwwhKDkKLqSfzqR+zV2AATG5f8ANT+tKVqsRNGT oxex6Tpy7AnMYcurqf1r4rTt4x7MVP8A8if1pSp95mFRijCxpi8pkvlyMnYo+yfFSf71I/Zi4/8A DJ/zp/WlKhYiaJdGLdz6nTE8/ejpHuWn9ajr0ZMfeUt9nhgbQFp4Y+NKVEq8pKzJjTjHdE1Om5qA PsVKwP30/rUliwLScuwnCfJ1P60pUvETZCpJcHp3TxUQW4q0jqPET+tQpekVS32FuRAtLSt2FLTx 7daUp7xK1idNEr6DljlG/wC9P61il6dlvxiExsOIwpv208/nSlS8TNjTRNTaZewZbUkkDgFJ59et fE2icF5ysDPLck/3pSo94mTkRnECeBgKX/2/rSlKrrS8InKf/9k= ------=_NextPart_000_0000_01C0CE5D.B524C4A0 Content-Type: application/octet-stream; name="oledata.mso" Content-Transfer-Encoding: base64 Content-ID: ADoAAHic7bt1UJxR1KeJN+7u7u5uHSwE9+CEYMEdEtzd3bqBIIHg7hqCNxIkuDceoPEAQ76Zqfpq /tjZqd2t2t2a31tP3a7ut+957z3n3r6n6jRkGm8TXE++Bfc/SBoOEe75BQ0O5T+9B//f+A/hwr1+ /vLy7+V/b//p5X/r/1N6euWf/xBffYf0CvIr/3wOeAX1FbRX0F/BeAXzFaxXsF/B+a8hAIf3Cv4r BK8QvkL0CvErJK+QvkL2CvkrFP8tTv63/t8lbTiX18sTjhZOAc75tXWH8/sft4L/QxG/Rsx/7wv+ f3Jv8PF00pupPvj/vF+Yw/HB8cIJwonBib62vHAir/D/L9gnhUOA/8/j+T/zHYRXpv9XjPxP9L9q //9u/V+yz40A55t/WoDT9dHAa9EcH8T0levlvWMpWJuOFiQYgq7IKy8LLmjh110BgKma1cj6sNmG ZLQqg6KDcD81s6fgGjvh/iUxVlFAqn7LS79q8b5Sjkye0FGIuDIBZQzz+exp7aqzt3FB+rHw8bw7 ZzSwceHLlIRPUcCN2EHPk9Sn0+abZyhVUtsjS+BRscZpRNHBmWmdz4Mqztrn6NNOOH+z2/HB25up opfAB1XzB+3Wx37IIZ+5g4zt8DQbNGx/um5Ox+PJ/SSIFSYPNZClQXl6CzWQ6801D+lWkqVBCyCE zUBqQwK5YHwQc8CTOFQBYo4TwAk7BpMAaQBPKFCXYGnqhwTIBmoAAiwGogrSCJUme8gDaYRIEz0I Ql0in4mhc+ANpG5CfzIYGcRWvpfswRg8HSpN8bACygIGScDEGs9v1tfPg83gnwogB3JFON34DwhQ DHAvxkMnmCe0h/SBC+ISQR0SSAFbAW/AP0lBsmST5M3Rn9Sh5xHP0lAfuV6KDpwAMtgepApUGxyI A/sMHgH10jwAQEG4DwGQ8ygzrCd4KAd4BFwb2kP+QA6lCX2mhSHBmiGfQDJIAVSwa8gncBDpgxIo iPBhBLQo9yIK05B9YYRhQWgKm58AF+devsQ8st21QNM5hh8vX7ietzvPZ5tk2lwXVE9OnzhlAsFZ T5EHqzcMBuqBN132MRz7VKN+bWm1habX/jIkZwVBhUFFI9S30O6zv9Mj170ot5crF5SJtV+elbb/ zqnmPGMlSL00/U25v0G+uupB9Xl5tNXw2rs6/+7/RDKxIvEkJhXk/Hy7EUgtKnOAt/PpJT/KZQqZ 5LKT5uWmF/MlYOSqxu8J56nPzO/KoOhprnZtd7exCENY3P+h9a4uYcX15tnz6VwkpGwDhbwbe6jd O+L5jh4WNMfCG3CeFwj4QSnwhPJ3d7GoaOqFPv/pEUKisXKRm+LU5pvwfC53deN8bPvryuXRNu3q hQ1aFPlMBi2K6MHzl4JNy71IwnhkX+hhapBepAABGA8wCOnBACSD/lQF6UV54oOaRz1jQTfCe5D9 EWEuchvo3XgPwtDXm7Ge3KCv84sNO5cPwvcXgJ1CegEBAFgBpBfnCQAJQn1SggRhdiM/aIBkkJ/s oTIRPRj+HDBb0AvmgzNUJqoH8MADOo/owXkIhPaufpC5rgW9kD+4gGiC11GeaGEb8i5RzxLQKNAL 9YMlVAlEAKWCJMm/IMLUodNyQdQPNGCasGcEmIzceWgg5UMR+AX5AQ2qAmUB+wDLIxelIC/oTy4Q DaAM4tM5xDz8GR8WJPvCAjODvoQ+88MaIS9YT5QwA9AL4KEK8gh6IXl4Ad0CZXCeEGCLIPPwwtDC yKkeVujfMh6eAyBPsDTqgwUUCDJHeXKHAsEk8r0ID7xQA/le9IdfkCggDXwAI0wLogTs3ZfsUW0q atw5GJnBkTn/qXa17usgpva95ly8yKXiYNJR1JD+7i8i72PZ7ydK7/EVjUd7mdPyp6aiwb85+dJr pSppzGrCKRJnvecfWnvPa3x6Lh4Muq1ehjf8d8UMKAOeMS5e9jQ+i++OuQQSP/fuXmh8GRhOmrzM jit7uZorl/5odnI3XiuvJ4r7ouEsDn+fQ/No6Qh09h08bgeU7ygqaDITDtZWoGdsBjl+2PmW1f3w +7jGdpTvctnYGWTxyCDSrRcq0ZdwWN66PoHQTK5po4/5shG0wnVlHlaRfT46+6vveMCbOJMtVmYs jePtHXO1/122wwjhyoKfQccSs1quIHjV63ZkZgRfJJzE1XsYwwwJ2yoPlxEQI1k/HYFMO+27nVan tg5fuQtyZQ63t03x/E0At8saHmy0S/tsnhBGlaGJeEwHcAmajrp4wXWmLWemEXu6T7Lpf+bsqiIK bAfh/03CkPNzvjIZziWZbDswi5ZRg920ZY5e7HCzFvlsEypgF7uyxnwYTZkTWSeVPhpVY9HEj7L9 sSc+uBw3WaN8ERtZuvTt33PmZJNBb8g3Ff72M0repGZdGlieuZEEfSA+KWdSxp7FtI4QlWTA6ZbC iI8yIL6MszXqZE8lvNyn2JBD02vvXvgAwZTVyQGPyVVi6SdKnCCzy195Q+fkjC8zKf7mqHc+VPHk Sd5vcXs4wHwQcBVKHMcSdd2xG+qCZIfU9PKmp/i7bZJqiqJqdaVbv/3ztx1KB53ih4RkBHi6EfNA omcAUohYKjeu5g/aJaVzVijazRbrQqS5fXVMpPg4KcCeyhJpMpJyqJ7FzG+yedFUgy030wJLo/zK RsZWWGwYesKTQePV35aq4wkp9z+eZOQPOU5qRiUknrJLzmAkY9fvg6JCK1RKRKNZet4+VQnbs4sU 2LE6zpai7en4vGOH0X2zo0awZlwTKLo2CT7nIyHKL66Dh+9hmG5HWFnaDPE0nRU6orSe7Jd4r2QW ESatb9hejG6rwqGLADW/bjDAodOj7KqS+VZRoKCaNlit43IQzMbCwpUXUrLaRIkvKf1znp/rXiGO kpIftt9a7OFIPP4uyTRD1MCceTt+d/OtiXFaWAfRq3/84DACwF47cazpHwiyw8AA7AdcMEa8HHkS 9Kp+y4KeeC1H5YMOvMDxsttU/7sa9I0okPkWrT3TUYhRW0g63BDvZbS4m63rLVKbFWoCvbXmxKef +pbBDPaoFwfGW5v+m5ub71M84/ja2N2Et2nZR2JCKWUCHz9Ld6U+iXquv8aHJGS3u+gR89F2O4z+ CFk82ajODpKGGOSZ0CGhi/CXT1YXD/wtUwiX+45mWPS3ssRMXsoKyCY+xaTHLKak6DNEGDXvPSlB 14GZk2aZG0NjQufZs6jfED97cA850baVPT6K5sNPBumx9bUT5+4qCiLk0HLipzUa81Fqzh3HcHgh rCQVX/cih+FD8eXgsCUBQjbdrndNc+OYIc1udSppgOqu47ki36uEcwgqCOWQ3X07YzHYqo9mDUxO vNviCAxONZkLGUhoYLHzDkHFLG/fRuaR3wPXnXRyNrRfKsEfE1XqDpQVq5jFD9F5sm4Z0LmQO89r Ha5/DsQKNe+X01WWyGW+gjiNBNK3rV9aYKj+ZTMQRS3bxc1l4vdw2beblqwThcavBYfFs1G9CyUV LRnX9ISAX/iE7Wd3N5Fxv6x0bgSdGq8jERYHUTOwJg0smvlIiV9Vmd04JMeBweRMhFh0cJRvv7nv 02pmb0RhfYl1+p2xcVnAA/zwFxdVjiCXDhrM2dDVREOhzlH4nmbvm3XlR8uuCl+D6hCWSylZar+R T2Bo2Ra/2wap1VA8PhzebWyoD2YkEXLYexsv31rUO+OMsD3D9Coe4/k0tmXkzLZDkcnLtosfw1qv F9dyGJCh1Ovy21Qo+LfoUukoovb+mzy3KNYhau8dGZH7r/SEbUWAby62WdS0yhh4NGgOopnCeMK0 4QhywMfovRi0L99+3C0zfRs7nq99nonc5jJXcU13Tu+Vh/ZcpNqjfsjjQcF3xcXdNd1+IeqvJ2Sp lTkWB8SZgJJafUHG1lVl7I62X+pYrY12LF4X/Q+0B+vN/U7zR3bPvDuaFVyvcsIzzsAhboOPFlM/ GT1p2O6iqQs9rwSRUBMHlCQpKQQTDQsff708qvhV220bjko0isbocyU5ve2mzmBrmVyZvgWYitbM c4zUBha7mhUMu9jjBq2OixmOcXMy7ByNe26O8Qln5rdxYDOvtPyLImaj9edRHARBGBAhFJocQY32 qP2iN6/jh+0+0zjyZ4bFZLURgIgbomqlZSZHPYV1IicliMJmXeS7lV6Ki+SYDTnJTKmxQr2ooMwk CU9WUDkrtU1dDDxh4vkYAaYmuvkN4xDRRLGhMi1QKvSI0JHuQoR2LEm+j97a67f8lrS6W+o17f1A YC7uqt83OXFvhBp5mF/MAVjyRIx20P6P2cFWWL8+1ICwlSdmRYN/lVFLw/qwCRPR8I3Jd1NRHWIZ 9FgrQSpWDaXojVjQ0TAK4oO9amejn1TbGQMhwo641kPotr1b8a6UwTshq5uBZb4WOIl76W64l3H1 1dLhI5Vebf15IKDgvQNrSaRKtFeGD23Hn8KrWBihqL/yzzHXzAy1i6geQeuPuFhoHO6x3AcWWcAt gWynkLBqYUY16NI+idNsqYqeNGyYB07QuQWVJv88Ys4ujX+9XiKAv/evgbjyHt4eBKyjgSx4kevz KKhFNSTtkyCBLxCKmNguZ6y3TSLUxELIfTb/bsdOsT4YShlvtO73l3wdyzCvgf/a886mGHHUZ9Yo 3oZiYO7iQDo2jGnRTnIsvxhZx+3G3lT4AGQjb8xmpkskjC1sfwmX1dr9+ATCVKBWb4e3Nxbz3O0o RUIdC58JLw7x9OwIDTNT1ATJldsgFcYOK7A+Bqo9nE6gc/lIVn5RNE7xIQviv0Cq3Y5Z9cXUL+Tc XtVdh+8V6FFDc7qS6KUGRBUr7lDHEU5SYoaPtaEh/+Dlzur3W9y0qNUDzu0Ivq54ExLZ5ExahRuS p5AhzT/nH5WKcJkRR9KY3wC1k5xVxJQVNJiw7fKukejDUSpefPg7zuudF5CV9pIxl8M6EM+sInY9 7IGUnuRjOdUck6gEgI914ymqSRXEj5oMH6lq++WIs7l/6d56eKvPjCZLiSIy3QhEFQdrnVwc/o5F 5qRLiSHNMQJW0uv8+JgblcEcO5oq0EPzYG73Gj7e7K1rPhsxBmZA8/wVZ8/aewQVsAHOk8ybEO7L oEgLLbRHD69HhApM2UytqCItZYZ0ghyUT4JLH4k6PmI4V8j23FjgOISnOFuvd9AD4ghvFmPoARdV gAmoU8h2SQLYclngRCMIExBXM3NyU43Fxp2hR3Bj+5PCPDChshdxI0/Q3h17PM/li9IZV+MBZM1z ILzcNy2MH7oadoQEezOAu7qdnu4jaJtVBY/OhNX182YT15OSZmooeMA1CIESUVDKhnljUDOjRjnT 0TXcbJbLdiLkfUTe5Zps+U0buYiF1KX+5J9yGfdNx9Hd/BbuwS8f9Gq+i98fOJ4OFHjwcgn9ej1D emj8Wz6uNieOwEqla3NE1KZLWjFlwrGvbIiTaBVO2Nhv7RJQsNmVZNkQT6Zj4osPIGotJ0eYJHsG uxGIUs0sSybfWJIoOgaFs23IoZjvGKFS8pERbsC6/mztW1JAma3DwUpDpJQo2Rb5Xl9+lvkoo4kt X/Sv0/4MrZXpfbvZ0tF2dl4UxN+UxODP0/uKvM2EL9N9w6o2x2qsdwI8P/Oyil1vyyS/eqLwByvJ gfsxPefY4pdHFQlqORbyV2KG366uYgx7Ej+PTSA/8D6t87PgSSBqA3VtXQdjP03wzV+OrKeeIEzi Dj1d8PQpY6sfkFPlUjJfFJN4N0yAx+CyL5OtFLxtcaZQ6KV29yEvyq/+aJDwOz+QDlMnWuOco4+O kRTct1GFdqwp5BzqND1SRpOFAnOCS8HGEYQzyYOhhf0jnL5oNSme3pkXqRmf+n6FD2oFRqPAU1Xm n6ZZsk/JadTbquS2TgQNUSjbBc+qIMEoGREtp1ANvs/usHZYxjE7tQBGOc0ZPrDuzhqMGL6/s+51 J+GagO/Sj9Xj54w9ZYul4S/7IrbQ9LoLH1R0na6d2+KMrq8nYGjL2tu1h3mxWFJ//FCr/pm3WipF Tf78gOcQnwOq5tWSSmL3zRa9IiIcr32wICfAGbNhmXWzsQmMUoy7sgopLzYm5u7blZx9ouunKjFx SfHorYvIpYtKMChN1qzoHHosOQyAQrgjh/54t7ePGamgfQqAfwPwZw5w/lNMWJ5JxV7tVU1YP0qQ IGersJ2/NRjHbMBdyt2z4OqRIGzv7ozRYb5v9Jnnm7QOcVQFOFCI9M0Z6Xw6Z2GElltCZBeaBh0j IEwo9mFt46d+itE2eavUkwzmgVi5oEglM1CnUY/OJQIH+ajskHMrQPaHl6/HJtcspZAkchaMiZKI XsIax/0LxLZRq/ns5OOk2buVBwROKzmxVeGJdL2P6mncOAMk5X/e9bV3pjKFAZplDuI5RyI+e+93 z/OPLdnwRwgLp/iZ/ajGSghTy82vGkAl+0h2d5fFyfbErmUWPGSa8rbDx7Zs37bKNZl1HslZMjwS zHK5ExSq8qworI2lX9W2gQeCdeAaF2McILuXd0Y3U2HTTUWbOv1kT8cjn/pKEoM7CFzRF1HTUrxk NLdKSzB/bjKDOvriXMwRjfhdUxkiNa16bs87Pq2q7T6vOFv6W7hacF/O2drmnBt98LqbEjfWXRr2 mF38huKo6hkLfNH5ZXLm8rIL2Hhsn3bUXnG6XYt6q0vH/gIh6a27Ft7hrDBqsMzLa7tLMD282GIL VwpiPR3xAZrvm3651ccXDDMhEBLky0pmORIDBAQR84/rdGUoGGKmuvGwFgQ2xbSNlE/doKmb/rS6 3RMf9jQVPKb84o4c5Sd32J2upWJbGWtPmnK8674lwx3PsZJ24NUo3sXVlrNvqdxw90U+LJnKr/3r j8f6DkO3ZDpgeFnWjI+FpeV14/wi4dTB6GkiLXpwDGrajAi/93DLYCiJx+LHkbHvpupZdUGGlLsn xCSBa2G3s9MfuP05WdKUsvh7rtpvuFwGRFgWvZ/vDzxiWfLdoK5fcK5CxCVvK+44sLFuyrY7Ih3m cHJ12/Pffj6SaSvqg6/xtB+t41fl7sPYsFf9o3nZvNPV7d2zp3atlNS4+R6CRsgDfzpSDny3+ufo r9GV95bqsbm9406eyf326Xvn17xx+bOHgsHNDfablKkt/HBpKaqqy8sIfw3qahmmt5s6+bDKxkPF UWmFAZP1jzR21wCXvClSlPM3RxZLkbQaOBlokvhvVxvmtrYLLFiA2iPu3b5ctkkvIaprTSaFevem epqa2ePq7UQO3ynBfrxZ8MVI9Dhdkns4NYQl/IzlMi9aY923f8lVCKtblPtynEQX2PO5/Jcw7uZ/ rlVP3cfzBWC4cbgbuvFD4UZgwndpciwt87/lX3QYLZ77aWgffsUbdHUYJA6QCKXjUK4e+yxWcHnu DxfV6ZeNkyZjPeG6qhpVF+Zrcw5Kf/HhYuv5EM6/RcUg409IOkcqyHBld1V1ZZuwIgRbdbsB1mvI eD8+nvkfUMwozOWDRfkbwuZ1cPKF+n09CdmD6vDZPVSC6J190oil6uVQreoNI5rNuTzzkcm2Cwg2 TPvp6Jk+Gwm1QMeHqiIMbnSwgYhIPq3o7Vs2JXlpA3akQ5yxPYgrTQChrxMnBnixEp/t8agGhpqH Y7vX0PEK9cAl7opBPAfbZQRzu4k1u3/7uv4OvU3dv5UmlXuh8UeDjXKDBfppDJDJ0mD6SYOLFa5f z8gpcWzlChKJPGNUJbQ9EYh6l5nL0N8bs0gErVJSsIf5e/9O/ZMY8rktqwen3zNj8wxZgFfm4UvH H1aRro7txhsbaGfUaZgIpszn48mi7j0UVAYsuDf8Mzpwkrv8tpRICB1cQJZSyeJCbq13IAhS6QPN QL1YeR6EOcVHMEiQUV3FTZjw5+eB2kKHrVsWXwkgq9bwX0yfvD2hidwP5mul14ZU6xzvq51Z/IOT PTRt0vnJFWMp1C41YgOw+yPI7D5p2x1YYB3ONXjo/bDxIeqXXB0ilawWJVt2hpJRv3TjtMMiMWJD k6ivT66ePnbakUAzqJm0AivkT6J8R0jNZnoPTHmoHdAYfQ8gpPkcmfVnHWZ5pbN8idVgVjhKaGMT qXO0M9uZ2f1AbWdgOZccxd8neA2kpDIzG+jMYq+yPwya8NTi4usFRslnan1jLtl0FayljHqzu5B3 OBRmSerIcV0TizOj7pOGT/OeHG3l6MfhPWYIzfVPMlQJbyOQArk4ktS5ixS/OOUa14TDUylZYr/H MbDqrP16MQbXqnR8RhXZxQKutZz7yIC8vIGJiLDUqf0dA8BekK3Lo2uiTIfBTd0r5igMtwmZjTJ0 znzt8uhmPFxyuaiAGRMbxaQYXz1g0LVgWOmYrN20EecSh4CKp7MEHT4sOvIujYz8G1BGoOdRO2Kd qXh/jA56/Lf0bpa4U8Jr8fo3o8wDhgtVh9YHWf3D+5lgPqm/gJ47Dkmqs1NIxHkH05QoeVL7uxnG Up+AvyW3JHR6bfe/7htrehdn5CWVAfYhYQUQSc+oQspy8IqPqmY2rr51JZEYVFXLWoMN0b7mZBYO GNmZpIoSLlYZlXyQTmKLCBusBb/D+2JpMruySVcdOCeZz5y2hldrMY97yc7QJFp3vFPOr1sQVijT lsrJKSowRGhDRmd3iC+eScdKl8zsctWHfJdIU/Qgs0d1+rdCnk40O2YCcTGaWcpr8z7sy1ZtVzcO GjZdu7yIfO2877dIdPg398ojPq/n6Iv4M3vKMLr0yjZFEUG9H3WGtNTbtNaM2ovOMb0fNptL7XJS RbeeXSWdf91ER/obrgihHYe6UPVI2iIAGitu53zpwQTZ9LJ3cGRqLN7D0jneAZBkSt3QQgf9W+rq OslIUkVeoGaV9HW/ZpRjxeSkdq13L43R7ueB5TgvYmNqBKs1Agp8FKqy9qkn7iWhrqa7RYtua2+t JBVvDxsyLGAxjQhU7RTrNMTOBjMSV7P8VKVoR1BXmXAqd4YB1K6istKcdouTGga9WRwk/swL+wwf AcU8Jpb7mmSVxUrYMFgeTk7r3LnF0+2Ltm4aljf8W765Ton9HDODHKZVj2kcKszzPYSW212TpGc4 kzApwwgdG1tqktIkVupTl2aMUgO6SK+45PFYyP389cXZ4NbmdxGA2+1hPSE/O6F7GTmZ0GOj5zBc Z6yK76jjKcv1VmJSCiVm4qrF2D2ofMWTfQdYlCgs0jzwqa9eW/0jyke8hrM7LKQJNawshs8CbwSJ 1ObVeoTPljh93wCzYxe4xNeL8pow/Nms9dIDJ1tKJd6v6eqC1Gmbg6MRfKg5Fxdw/+ZguvXDi1hH 8hKbDkMyiFIIccMK0U3dLrreWJK2Xeg8ipIjBw0AWXn2EXZ1cBL8iJCzf7KMoSH3Dx+ICPjM2wXf /lDAppZVHaifkGBdX3LEL/4D0enDygInCKcN5CjdXfCMMhuQHC0aE+SmFD0LFF2vht0b6+RPOPDv 0IqLzGq2WKKyzdCSSFifyRlMmAizrEHQ1SYw/tT/hFQB0BY0Z8QOs7l/+2Dnbk9p9tW3uchx5ny3 lIXUcqy4Ld/nH+vy5azgt6tc1Us6pNWqCaIguslbx5WMaZKWfORcpiMGvTRy1G2CIlaRCom7Shrb dJ7H/Oyx9zUcneDZYgMMUjCavsmrnOP011rZGjY1xhbgdw5Z2Cx0fiKsQdq1xqGekkjNBqbmJQgf 9NH1Gli+E9hQFNSrIx3t/97KSBn3aFJ/5uF6COGA/Mw8SP1kbno3AJUIlsg/XchYDxVOSVp77mk+ lSunC2+xJN26ZPQRgyZ6YzxK1xjHZoldvu9eBNrxA6iQqNkPPSk+CETSyUQpJ2/aTKhE0FnrZl95 +EXA25q1o6ALVeTykTsqZuK4vXkUHlH/g+8yTO157DJFciVr06FQ9X3cDUL9O14B22E63fPJGpiw eh3td/3xPXkY0kSX9JWYYfw70qPOSH8WLwaobWzB1jcsWQ2i9VmONa+15rwrrLgs2biJr6msIm9m NYFtgueO4z+mvBY9LgmSA3alVruYMnwwm8qp78nW+X70IhzOAyNX9g0OhBrfq71JHuHLmsKpI7qf p0q7ju8Plkd3AorgCAGFJLaeluO5Rzp2JcjCLKfI1HLot3cveujUqTiQp60ZZu6LEzZlPG0iZgbp fPdCZFxaCHiXYpwNsIUkWiEaR66Y53DtlnlLYR7buNZwBFv0C6WMqAe6JkYxZHFrbesqtauiFfBv klrR8gsYrYjJJlE0plK/esV1mzwlQ65EReLO5vydhn8m8LoSdB5zDyTT1QIDvOo7qTuOMvbV2XLZ 6TywpIRzuWbaVrLJf2wMThrjr/O9Gf0KDNyiQvRxZMJWtynFMM/+U7fQI06Dp8hk6Zyjafs07GLs vrhGYz7C8/4yRVvQk5Xc36ZIgiVpaT53IvWq9LM6j0dTHqxatPjABFL4gU1WV9PV8WM1kkx51ANl M1KfxQhyedUU6cmCI+aY4Ry1hsLEvlNLhADg8It3cMfy8WJnjBRUfWbUEeXE/uStyIOwBh9pqC4J DoE2ddMc2WvOh3wp7jnrTil/uvqJPa3uU/aAjX99tW0L4WizYKEJVXrS8zuUk86vYNThkj8NR4pG pEzWPqYTbhja8WicueHrnQsH/wYeUjpM9p5RIU1TaDcJ4obVniXiOAqRUOTjE1it5z3JG+f0EsWI sFT2NzYQyUkwPcJ1JDxd+ONch9G+8/wovGcxqjwPoSuxc98PHRxQKx3vzEASUom7vV/uF911nAj7 pN5Kj4RYSni8PYnR5I4t7pO4ZCLYHnghGN4cnv6l3NAb40AwuaTFOaJ1DHyS1XXjYzOVO4WZL++a qbA0Pn9MEKvENp/22F+ZQQISYXB775F78tFZ8gZRR6XPrwu6fE7O1UEQBsfkZ/F1yW1vM5NzmK7x K/OWZmLeB27PxQyB4hUfslwJhawiCkqHio0fxfWDXS1fFE4G49R+dzEN+8TTAwbbByQUL+gW1AZU SESaSx4x3zwI8gO5YV8TVmOred+0iZOJ2edFYid95f1pMXbqg1H9mVEFzq5uLbFMlpAGGYOyWpQr jnvMIiKiTLxscGsZ4+hZT41qZ/8UL4/gh8zkDBhbIqgS16zxs3AfzE7GxaqLk954le0tQ9WViIOw wK3UvmvWuivd1W3cSZ6eWdOz9cPhG58yAgWEqb+tcfYbU75GY6CM0iW3hHqMWDiqbznY02DKVk3j GUaT3GZTnywiivxPFbuMiZVy9K3UD381D/a50H6w0ejz8REkgS4Ttbq+agSMzY1ozPZcP1bnIB63 nPaHjzn2C4lP6TJi0VliN60aAHAKx0TRC5wTqYsIm+wUbI75QWN8HRtJP5Zbq/vAbkD9GoGHeT7L MxLb/KLfB2i342hl9TgI1MfFSYKVuZ1BYUBfmT2cRJJ2kYOSEJECkBQr+dg7j7DxSZFTc0ySGwDA Iz8+KmPsgaND8eo54vD3kwwNn6B/4aj2Sc+Pif6f18is2zYzOyLMWlWifbpoPFsdRArH9qkgo67M r2KL18P+kQ0yP3zqPjj8qUI1OpBwQNvgK7yvPiAWEz1MqtGnz9U1tvEh5o59ayT+VvT5SHTLmCoU ICAtCv7k2X0So6epzKLH94t+qI/QNHGKuTU4dIbQkACS41Ql6c6iMz9Oech8ypnraxk5zQuXa6ul 07iHwSXch4PGbI0ajDc3WLejRTGCybWRHzZkzoP+URhfJFJtBiQwN0GVB47kHTZyDcH+tYlE2Fn+ 91f6pm0XYWftV9bdP/mKZtbkhBLcRdxlrKIG0uxDVeFn5AE2+6CO5rc+akYO+GPm5dzisFnVR1vh SA5j+xkNqp5UFWVLkHIgN0ULY+8HQgKB+2iOc9LhCpb9nAUtqI61rczBbwOAjufhKvZdIh8MBrfA asLi4Ol1YZ+jMEgpSeDv0jqeeFNOO77j0y+PWWZ+88VmbnTkPJTCe4MZ1WsBVbtfG+NEYraIeAX+ SN+yMNF4JFL4ACwr1Nwms/r1uof9yrJnF1uvYuILbp2LE5Yb7/S6Uf0lguA5J1mMX1DKb4dTo/5+ q2TEqCr2P+jHJG2P9aj6ia56Joaqc9wIorEl9tugD85MacnrMuhx90e6ifP0F2QlrHJ5MxSPuW8C VeOcramMHanRd4UIWlfwe5Cx3Fy1A0ZI/TEGFkus0iIVFAeC15gNRNIZxwGnGjdw+KzYA4xde5HJ LhVk9nYUNscTbsTiHmCBO5Y1bY+NykUdmG1HEkxZROQ6UKbmx2HZ78SVAUqlR52FCnetJE+MKs52 knkx33lFRmB6nLlXi1NcNKb+AZ5xlA9TjoSABHYsz2oOAMDUU/d55lc763Ir4FhoQsN2zqfoBJim VTLnPRVJC9yZRsdWj+wu+fyVyZ3V5OonA/F8AFaIl0vLmZMivj/vCc9DLTsZADOHdHuzNKl9279z OZANn2BY8k9Pz1t5PkfEgGrs24+FmfvXJOU8rD+KHFuZFLDRrbeoBv1yG+m4m2PuYm1IXGqDUi6T z34biAxpmsV51FEzdnEt37Nu8ZLVg4cL3YPMwksozIdnWjHxalnShoHcBUaPmbJ0E/ftu2CsA3RM 92Oedy3CsKpHdap05/yzck1tlgaKcnDfPiWmPt0pZ5KW2C/zbmWLywLn5usb477yN2ExqlXXT+gZ RoahloJXnClWSWAvB4gQOM2yRiQ8sdoi0e9WKJ1hWjoKY26O5qvCRBw3py3HeBR6HOVu7K70vrIM +yfZx5kdNwmfOl5X+QZhJ73r2abgpULJajEprQOHNAsckvxiXEq7H5pYgRZuDxIiZi2/UnFFczXe eu+WBxM/ECINt9IWJ7dtlRyQoJ2C8Q5sYa0pWLXcvfauMjvVrkdmyX9CssM70Smp0IbIsITsSFcz nCFUp/pPHigxnlu4t6woeawQfeTQFwhXf7Dqre1D5ZWmxRJZ7oFNdTmrjYvDnS6fWD/85yGbnzPI kfyfPOpQ09meoeuU7W1/+5WjbFIeTvY2cdE9LhsCEdpD+YTCB8DNA3E3Ftl6nysMWWuD/c5NvehM EAoJ4q3vIk3kn8PNcqk4LwpSUrIKsNy+TzBpCX7oiXWNi82qXmn4nslpKhBaPl7FeU81F79MREgc JL+z57z292igVfxL0WCaVPDSTKWKL9z6DDvWoD8x7vk0YuGqhSfBhJqArdJNAXUzZjR9+dBrykz8 k4ng1wf3v4TrjK00iL0NTtnD55HvdH2u/K8u7zYjbzaNxvqLyT57u7miRljurYJ7jmcF3wvO03Cf 9jOewVHOmUbJdQeLndc0+59kYwvPA3k4VgRSS0M+CkpF/qR2YH0/dla9bqrKDxjp+vAORbIjC7sL lci4xufb67y/gUemjCo3n6ZYy6R81zeBL2LP4tN9lfBOZhcK/oyM+BzJ2bCpOvEYNt6kXMEgrOfF q0rExVA1N2g0LG3kZDadQkieW2J3qXQmJLVQxCPYJ+4hOYTw/RbLaIXn/ar5bRy/E2+H+GPgzNfy EnPYah2AZZC1a147WyxXoGlTi9Sc4gine+Dt7fFO2NI+XYaYwrAcnODiie2boBpzDqw067TkK1US iiIcnGFRLXb/Q/A4jTiv+WFp3qG/EhcaZb5nY01EOWhMqCfmQC8IFu9JLl65Xinc27GZ+6fxIJLq nfgWXUkd0pvpRk1DVdJGjD+XM93FvfNq7Zv+198oZZ2N240vDbMGz8LHStL6DTnceSki+1OvzeM+ VNKV/XTSnoZfzg/rjEvLOXNx49er7PtUuSuXCXaN+Jrq0ZnZ8o26iuBtTUBC14QQPo1JG/nm7ygJ AA6ISZUllcrfIc+EIaXp/tf3+fiwHLP9ZAA6AVJBJJKVcBBq56akJ3akMqnSVvK4oK6eIBeVK3z1 8trRAZc4rxTpDzFSRgmvgF5AZaRxPRc29r7aBEggXQSs8qtpfybxkpn56YnOxcpyolDKjwDIT9gh 0zPoaRT2/X7yi73LfoHhk2eFgz361uz4KqX1p0bfd+79xMbbSV6HEorRl6aPzd8K0ztmf6kpWsdH XnZmFr3smefbiDiSODEot7Lqm+6gsaaL7Rf2SYK8h9xXxMLU8FhzJBrOv0iqf1k+L1+KChE7Rfvo Kc6sRnmq4UMgCjWDWJH3/Hj2zKQsfHuBLPHlxX8jIa1zugTnBnTzuLqWFhipYHmWWxPpQZvUw/3h hDn9ovBbz+YUaRd7g5Tq09JDcaucMN9npe0S7jrPhQ9u9bXr9qr6DZexkxRX0PvVOJa1P0eGo8mf 6AVPLExK0sjQW8kEdODN8SLGx0xevhkg7Rxnq9UdompOdbLg6UYirMnDwiM2V1Y1jD4vf+8oIv5e UdcAXeR6H9FPi1XmNpMtZQXMHzMzOY9z46/uy7MhK0j9Klibu6BOLlObq884oT6v8stdMmtfj4Lt hGXZtM1uFK1GWJ1sx6iYh3v+B/oxB+qfPjq0TNQVScaK2IIr/4Hx6tZy7hMXA6lDyAOqhJOKb9lz W1zDUcntwZmGGknn/S+sgJwrD5OO3THijxJbJRx4rvl2YOBvH8vdqZR+6aLtxDc4+qPOsxxOQhXJ 5bhWH+S5e0dqov0tyZhpIISAqyMKJUc7fkEhUZeC8pEHggwOYxmDuKSFR33O5HmWzS+/YgIr4mXT ynuxfzJoD9KHedR77Y7vW1/2R+7KzUs2IW2FPAkJ2c/ucmZwGUZIbFqAwCwtAPJv07f3xrE3oKRK MfEkT4E4dNDZ6er5NEuJlTOLzUQl1lRtsZ5h+GaGMHIrY6dlQRgdZoZB603rtnXQGhOZ0ftFl5/v MI657wsvthWt3Hy11LW+dxKiADS8rsJFsIXtZ63j4SLyIidPMCMI9ynGdNiB1dA/lPLgTgKq3G9c gAPRS8oyZ0QUxD/KMro/ZK4vRiC/eaPJ8DFnfttKXqjcyTyaAluQkSPYPgUxqZd5+QaY9rZvpGaa kvxoF/1mzPcHJnWrPiCv2kiEiah4QJAZmT9k1JzT4nOTMOR380bp6L7W9pd4l1quoBjQWocwkQx3 FUU8w57UsYY3XASpDJOcKHkygwURR4v7TFrKQcf2UalfVkI8j3Rkju/vfMgn66YvLeG0e2+N3UZP pQOPVvI8koUNWup+gob7NiBWFwGttqT0nFM85OCYEl+LoNHPCfIOo4+4owI+sYXGYWH/rML9S4IW pjlW88MGM3hDCV90+j55+oPAOxMjO6t0bGY/0zkSK9FEkLCegwaGk7CgbrdsTtw5juL6U+8OUh4d 10y8Cs98dEZHLas+dFPP6LN8H2feAOA5WZJOvfseqrJ0kB4vOh2NM191nqJltFvNqfR19pZUbeUP p1F8j0oasy2omkZMB7DRsRNmSW0zatsZrJ2UzIZi7fAS2IrbFVXW9rSH1L57oREeHIkrzrh2C3W8 LDH8hZshyclSSoMyzr14okwF7PvuWqcsO/mwHXwl6LcckyimEc3qzjs0KPpOpYn3d8X18cfi1MjY Vnv1GWa13GouD5rD709vehEPLJBDCQUvI0W61p1YSWyVUE9gcuJYNVj9yIKcHZRUWfB6RLyVfv0N VZRChf7GB5v55gURe0gmQw7eC2Rhxla670T8SfMrQsLc02LLsee+zNO/mz+ZM1txnG/4Vxu2kNye NL+WEWWNE+uvwIwQwWLml1DCm1SJx+RetJ9xTAcPInPasRRMDXPAPx0pl947xE/x8VhxHawNgQz2 mUapT98uJRzPcKmo26cn213JdSMAAKwqEp6zatzfCMAtyNfxkVeVLXl8S6UqaVyjxNvKImTagtgO rrwluE9bfMYg12js3xrk/qyqTFPWUnHV7jd1c2nSlBEh2USjvl7LfzA2uSdl+3TW3ofMJSS6STNn qymkzBDQ7DOpRpfOLFI4weCHNvGEOvUU98AvyP8WJCUuaSSgSTetyIo0Aj60/LydxJuDxRKDDQWY /P4I1dRY0SoukxZLm7x8oNVKKSq/Wy6TeOUvBi2+GPxnWI6RbfdPYtXiLOQsPxJOicuZSpIR4kdc a3QUe2Xb6t9z8f2vfnE0tSDlXi0FMD+FDNBugzSVq3h9lcLy0qsvaKeFlA0uPNgzf+igSoUhUtix +weS/sZtyi/JB9HdRg2TNffpGOrECJUEk1x58bX8lMOhxEKKh+LV2SSOiQ2LS0f2zShgNSKMRSdh a/86BJePAquw9jZpHszVXt00JRVi6Qup0o2XtZxhl3RorPKTZX7GufKKwoTFtxdPX10G7paPVPKT Rx1jlKjQ5kIVL2x0BqiTYrzQrnHLyFmGqWj1IrR+6hd+hiVOCfvTCUYOmF2CQvXeIka/Zc/3vPqG 3PrN4JdHjjduvtK4ZLuxf2VR/8y/WtHOZJyACwzQVibJN9KQAXrA60gjM/Z9L8TkgbMPgfZrsVZn Zti97avyUlLjZgQB49UcLZJJLrGR8NHt1tJ1hjHnGwseNNaaNIkY73NYPJwrE0wGqN90HqxTvxP6 Da3ivKZj5fkU9h8FOYrYzpA64BYuANEEouLANSCpFzeoGqjlxpXMNyG507vPBoz6c8A1jVeHpC+b 7c39esp85zPU/5XVs+hAJc/CKeAQFSqURbcUifyvPILUPZlOsD/VvUFzlREWkRyrExbpO8Nr3zJ9 n8NSdGHUfPjCCF8Q4CfBomDzlyikDf/Ze6RfRR6cZ3FlRzYNhv4AhDmJ6l9uNQrMqpuELzOi/quE md3djBBcvW+zVBI/sADRBe1BSQ3ease+iTalGxAcjWdZfMzFk15WzXbWRbGgNU0J80tvVK+nn1ck xoatnjVyg+qsMbk55eIyxX59xxdtOsPXwe7uYmi2wRdG/df3RPLarKtoeU+270805Y9wD5sgfcR0 tfLgd8je/ipUNVIVZBWT0tMWD1hyvKC6qTY9ujYdbPXZ1NA5aLrzknN9xxBniWWPIR2ZyIrHLDbO Ng3XAishvCqzpevgf+0fM4XKxjDd5n2V64gWKbEW/h1Sf5dMdfqbv2zPc2jZIxwjQFc2wGLa2mvS +J7XSpriXZ9+69DsAGY+WYAJbxZBr+b+8dV7MoVfxGry60CceHtcx9nS0RxmCbfMMq0BcnTbvZgB eiRk0bMJNOy3UiAIHOXl5IKYIRrZmji8jxOO9tkJi87E98ZEozZFLoBGhRtOikg5db/IIP0oq6gD tkhgR4Va+xVx3m/UfxuFgEmLDTW2bB2l8AelUXbCk6ZQsIRYFGrelt5qDaSu8LjeizByzOfBdTB8 6QqZ1m7B684Y9/sHK16Uz19OsjCjio4a7vJ4/lU/2BbwHrsy0JaVhsmhRup4flrWwOqooT7Vyj8g 5eAJroaWTrJW4/6Bqx4wlVkmnZuddN9HKehjMf7sjJMLLYkonw+GIsGQC7Gf4v6jbrpqbMC8AwWX S8ZqdNPisM+RnVtNtbpTJKpNX+6gMgRt0KdfrvgPXOHhKD4Ww4y4Emtg95MUTaQv/CnazeaXi+ht E6fXHzJ790yMkLbcCDtmy36M91ubDehZxa6Yz0esnIHkmm0YtNRLiF7kkX2kPuFWFNMFzOCkH+62 PO7MEZOCAv+q6g6+uWK6VadUrIZjASi/t9BhOFcyeNpZ1GBP0hP7WFNQ/ynBzMHrZoUY3YQ/ygwq JgWYWh9WLherJwT/K3oNjo4fCsue7Ozo2SJFdLWPH1NersyDaRbujPb8IScZxSPKCUn1GfzUdlXX m/Mfz/oHgcFtDpLLDu+xqba/Rv7Bb0V+qyW4rXWu6AVnqJNz7P+Bf9f8/1v/BQUi0ew= ------=_NextPart_000_0000_01C0CE5D.B524C4A0-- From owner-reliable_computing [at] interval [dot] usl.edu Thu Apr 26 11:54:46 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id LAA03623 for reliable_computing-outgoing; Thu, 26 Apr 2001 11:54:46 -0500 (CDT) Received: from vitoria.upf.tche.br (vitoria.upf.tche.br [200.17.166.1]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id LAA03618 for ; Thu, 26 Apr 2001 11:54:40 -0500 (CDT) Received: from holbig (host-2-20 [10.0.2.20]) by vitoria.upf.tche.br (8.10.1/8.10.1) with SMTP id f3QGV2412455 for ; Thu, 26 Apr 2001 13:31:04 -0300 From: =?iso-8859-1?Q?Carlos_Amaral_H=F6lbig?= To: Subject: Reliable computing in the high performance environment Date: Thu, 26 Apr 2001 13:31:46 -0300 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook IMO, Build 9.0.2416 (9.0.2910.0) X-Mimeole: Produced By Microsoft MimeOLE V5.50.4133.2400 Importance: Normal Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Dear interval researchers, I am working with Prof. Dalcidio Claudio and we are looking for works and libries which works with reliable computing in the high performance environment (cluster environment, use interval in MPI or PVM)). I'm looking for references about this topics, researches groups and interval libraries (projects similar to INTLAB, PROFIL, XSC-languagens, FORTE Fortran HPC, VLSI Coprocessor for Accurate Vector Arithmetic, etc...). Best regards, Carlos Holbig _____________________________________________ Prof. Msc. Carlos Amaral Hölbig Passo Fundo University - Brazil E-mail: holbig [at] upf [dot] tche.br http://vitoria.upf.tche.br/~holbig From owner-reliable_computing [at] interval [dot] usl.edu Fri Apr 27 18:58:01 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id SAA05678 for reliable_computing-outgoing; Fri, 27 Apr 2001 18:58:01 -0500 (CDT) Received: from cs.utep.edu (mail.cs.utep.edu [129.108.5.3]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id SAA05673 for ; Fri, 27 Apr 2001 18:57:56 -0500 (CDT) Received: from earth (earth [129.108.5.21]) by cs.utep.edu (8.11.3/8.11.3) with SMTP id f3RNvqX14691 for ; Fri, 27 Apr 2001 17:57:52 -0600 (MDT) Message-Id: <200104272357.f3RNvqX14691 [at] cs [dot] utep.edu> Date: Fri, 27 Apr 2001 17:57:52 -0600 (MDT) From: Vladik Kreinovich Reply-To: Vladik Kreinovich Subject: opportunity ? To: reliable_computing [at] interval [dot] louisiana.edu MIME-Version: 1.0 Content-Type: TEXT/plain; charset=us-ascii Content-MD5: UUik10O46Cv1ZDaIw2KJbg== X-Mailer: dtmail 1.3.0 @(#)CDE Version 1.4 SunOS 5.8 sun4u sparc Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk "NASA Explores Future of Software" Wall Street Journal (04/26/01) P. A1; Wessel, David NASA administrator Daniel Goldin is planning to collaborate with universities and businesses to develop more reliable software at the Ames Research Center in Silicon Valley. Along with the military, NASA has historically led in software development for its space projects. As business software matured, the focus of NASA and the corporate sector diverged more clearly, but Goldin believes that they may be drawing near again. Henry McDonald, the NASA physicist in charge of the Ames center, says that applications need to be found that support both business and NASA functions. He says that both large financial firms and NASA need more reliable software to manage reams of digital information flowing continuously. Goldin is aware that NASA has suffered software-wise in recent years, and that better programs could help prevent disasters like the crash of a Mars Polar Lander in late 1999. NASA has already succeeded in signing on one research partner in IBM, who shares the view that such collaborative research can improve both the process and the end product. From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 14:06:00 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id OAA07737 for reliable_computing-outgoing; Sun, 29 Apr 2001 14:06:00 -0500 (CDT) Received: from femail12.sdc1.sfba.home.com (femail12.sdc1.sfba.home.com [24.0.95.108]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id OAA07732 for ; Sun, 29 Apr 2001 14:05:55 -0500 (CDT) Received: from beryl ([65.4.134.233]) by femail12.sdc1.sfba.home.com (InterMail vM.4.01.03.20 201-229-121-120-20010223) with SMTP id <20010429190549.YREI24113.femail12.sdc1.sfba.home.com@beryl>; Sun, 29 Apr 2001 12:05:49 -0700 Message-ID: <008401c0d0e0$eb8aae80$0d0da8c0@beryl> From: "Eldar Musayev \(usa.net\)" To: "Vladik Kreinovich" , References: <200104272357.f3RNvqX14691 [at] cs [dot] utep.edu> Subject: Re: opportunity ? Date: Sun, 29 Apr 2001 12:16:40 -0700 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Generally, no. NASA problems are usually not with the floating-point arithmetic, but with generic software bugs, which call for a new level of theoritization in Computer Science about how to write reliable (no generic bugs, not a computation per se) programs, how to "control those stupid programmers without doing the job for them" (which is, luckily for researchers, anyway impossible both ways), with grants for researchers in these areas, with new Rational (and probably Irrational too) Roses etc. And yes, it's a lot of sarcasm about such techniques, though not all of them. I think that as a direct result, you may expect a burst of publications on XP programming, so if you can link your research with XP programming, it may be a reasonable shot. There could be some other prospective areas, but they seems to be even farther from interval arithmetic. So otherwise, chances seems to be dim. -- Eldar ----- Original Message ----- From: "Vladik Kreinovich" To: Sent: Friday, April 27, 2001 4:57 PM Subject: opportunity ? > > "NASA Explores Future of Software" > Wall Street Journal (04/26/01) P. A1; Wessel, David > > NASA administrator Daniel Goldin is planning to collaborate with > universities and businesses to develop more reliable > software at the Ames Research Center in Silicon Valley. Along with the > military, NASA has historically led in software > development for its space projects. As business software matured, the > focus of NASA and the corporate sector > diverged more clearly, but Goldin believes that they may be drawing near > again. Henry McDonald, the NASA physicist > in charge of the Ames center, says that applications need to be found > that support both business and NASA > functions. He says that both large financial firms and NASA need more > reliable software to manage reams of digital > information flowing continuously. Goldin is aware that NASA has suffered > software-wise in recent years, and that > better programs could help prevent disasters like the crash of a Mars > Polar Lander in late 1999. NASA has already > succeeded in signing on one research partner in IBM, who shares the view > that such collaborative research can > improve both the process and the end product. > > From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 14:35:31 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id OAA08035 for reliable_computing-outgoing; Sun, 29 Apr 2001 14:35:31 -0500 (CDT) Received: from cs.utep.edu (mail.cs.utep.edu [129.108.5.3]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id OAA08030 for ; Sun, 29 Apr 2001 14:35:27 -0500 (CDT) Received: from earth (earth [129.108.5.21]) by cs.utep.edu (8.11.3/8.11.3) with SMTP id f3TJZLd22393; Sun, 29 Apr 2001 13:35:21 -0600 (MDT) Message-Id: <200104291935.f3TJZLd22393 [at] cs [dot] utep.edu> Date: Sun, 29 Apr 2001 13:35:21 -0600 (MDT) From: Vladik Kreinovich Reply-To: Vladik Kreinovich Subject: Re: opportunity ? To: vladik [at] cs [dot] utep.edu, reliable_computing [at] interval [dot] louisiana.edu, Musayev [at] usa [dot] net MIME-Version: 1.0 Content-Type: TEXT/plain; charset=us-ascii Content-MD5: XUaEth+JcVv27h3i4TdxZg== X-Mailer: dtmail 1.3.0 @(#)CDE Version 1.4 SunOS 5.8 sun4u sparc Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Dear Friends, If I may, let me rephrase Eldar's comments in a more positive way. Eldar is absolutely right that NASA folks have close to 0 knowledge about intervals, so if we expect a shower of grants to come our way without us doing anything, we should not. On the other hand, I still believe that because of the emphasis, we have a window of opportunity to argue that the inaccuracy of input data should be properly accounted for to make a software really reliable. Eldar is absolutely correct that this has not been recognized as a problem in major NASA applications, but it does not mean it is not a problem: from my side as a participant in analyzing satellite images from the viewpoint o extracting geophysical information the inaccuracy of input data is a serious problem. Let me also clarify one more reason wht Eldar's pessimistic reaction is not inconsistent with my optimistic one. (I think we are saying the same thing but in slightly different terms.) Interval computations can take care of two types of uncertainty: * uncertainty caused by rounding errors, and * interval uncertainty of the input data (coming from measurements with unknown error probability distribution but known upper bounds on the measurement error). In numerical mathematics, errors of the first type are often the main problem, which explains the widely spread emphasis on interval computations as a better alternative to floating-point arithmetic. I agree with Eldar 100% that this is not the main source of NASA problems ("NASA problems are usually not with the floating-point arithmetic"). In NASA applications, the main source of errors is the errors of the second type, and here, the situation when we do not know the error probability distribution is even more widely spread than in other areas. After all, to get a probability distribution for errors, we usually compare the measurement results with results obtained by using some more accuracte (standard) measuring instrument. However, for Hubble telescope (and, for that reason, for a satellite which sends us photos of Earth), there is no standard telescope floating nearby which can be used to calibrate the original instrument. At best, we get upper bounds on the measurement errors, and when you have upper bounds D on the error, every measurement results in an interval [x-D,x+D] of possible values. When we process the data, we transform the measured values x1,...,xn into some value y=f(x1,...,xn), for some function f given usually implicitly by a program. If we only know intervals for the inputs, then the problem of computing the interval for the output becomes a typical problem of interval computations. Do NASA people know it? Most don't, Eldar is right about it. However, with the emphasis on reliability of software, we have a better chance to get their attention and their ears. This is why I view it as an opportunity to succeed (but not a guarantee that we will succeed). When a few years ago, Dan Goldin visited El Paso and our NASA-sponsored center, I was impressed by the fact that he did realize the necessity to use intelligent methods. I know that this realization is the result of a hard education work, persistence, persistence, and persistence. I did give my spiel about intervals then, but of course, it is not enough. We need to use every effort to educate NASA folks and maybe, they will start understanding the need for interval computations as well. The best way to educate them is to show that it works on their real-life problems. I think control people have a good chance of using interval-based robust control, etc. With the new worry, we have a slightly better chance to get our foot in. Vladik > From: "Eldar Musayev \(usa.net\)" > To: "Vladik Kreinovich" , > Subject: Re: opportunity ? > Date: Sun, 29 Apr 2001 12:16:40 -0700 > MIME-Version: 1.0 > Content-Transfer-Encoding: 7bit > X-Priority: 3 > X-MSMail-Priority: Normal > X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 > > Generally, no. > > NASA problems are usually not with the floating-point arithmetic, but with > generic software bugs, which call for a new level of theoritization in > Computer Science about how to write reliable (no generic bugs, not a > computation per se) programs, how to "control those stupid programmers > without doing the job for them" (which is, luckily for researchers, anyway > impossible both ways), with grants for researchers in these areas, with new > Rational (and probably Irrational too) Roses etc. And yes, it's a lot of > sarcasm about such techniques, though not all of them. I think that as a > direct result, you may expect a burst of publications on XP programming, so > if you can link your research with XP programming, it may be a reasonable > shot. There could be some other prospective areas, but they seems to be even > farther from interval arithmetic. So otherwise, chances seems to be dim. > > -- Eldar > > > ----- Original Message ----- > From: "Vladik Kreinovich" > To: > Sent: Friday, April 27, 2001 4:57 PM > Subject: opportunity ? > > > > > > "NASA Explores Future of Software" > > Wall Street Journal (04/26/01) P. A1; Wessel, David > > > > NASA administrator Daniel Goldin is planning to collaborate with > > universities and businesses to develop more reliable > > software at the Ames Research Center in Silicon Valley. Along with > the > > military, NASA has historically led in software > > development for its space projects. As business software matured, > the > > focus of NASA and the corporate sector > > diverged more clearly, but Goldin believes that they may be drawing > near > > again. Henry McDonald, the NASA physicist > > in charge of the Ames center, says that applications need to be > found > > that support both business and NASA > > functions. He says that both large financial firms and NASA need > more > > reliable software to manage reams of digital > > information flowing continuously. Goldin is aware that NASA has > suffered > > software-wise in recent years, and that > > better programs could help prevent disasters like the crash of a > Mars > > Polar Lander in late 1999. NASA has already > > succeeded in signing on one research partner in IBM, who shares the > view > > that such collaborative research can > > improve both the process and the end product. > > > > From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 16:01:37 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id QAA08376 for reliable_computing-outgoing; Sun, 29 Apr 2001 16:01:37 -0500 (CDT) Received: from nwcst322.netaddress.usa.net (nwcst322.netaddress.usa.net [204.68.23.67]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with SMTP id QAA08371 for ; Sun, 29 Apr 2001 16:01:33 -0500 (CDT) From: ljiw [at] usa [dot] net Received: (qmail 29331 invoked by uid 60001); 29 Apr 2001 21:01:31 -0000 Message-ID: <20010429210131.29330.qmail [at] nwcst322 [dot] netaddress.usa.net> Received: from 204.68.23.67 by nwcst322 for [24.128.176.15] via web-mailer(34FM.0700.17C.01) on Sun Apr 29 21:01:31 GMT 2001 Date: 29 Apr 2001 17:01:31 EDT To: "Eldar Musayev \(usa.net\)" , "Vladik Kreinovich" , Subject: Re: [Re: opportunity ?] X-Mailer: USANET web-mailer (34FM.0700.17C.01) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by interval.usl.edu id QAA08372 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk >From this we conclude that NASA is not interested in avoiding disasters such as the Hubble Space Telescope? "Eldar Musayev \(usa.net\)" wrote: > Generally, no. > > NASA problems are usually not with the floating-point arithmetic, but with > generic software bugs, which call for a new level of theoritization in > Computer Science about how to write reliable (no generic bugs, not a > computation per se) programs, how to "control those stupid programmers > without doing the job for them" (which is, luckily for researchers, anyway > impossible both ways), with grants for researchers in these areas, with new > Rational (and probably Irrational too) Roses etc. And yes, it's a lot of > sarcasm about such techniques, though not all of them. I think that as a > direct result, you may expect a burst of publications on XP programming, so > if you can link your research with XP programming, it may be a reasonable > shot. There could be some other prospective areas, but they seems to be even > farther from interval arithmetic. So otherwise, chances seems to be dim. > > -- Eldar > > > ----- Original Message ----- > From: "Vladik Kreinovich" > To: > Sent: Friday, April 27, 2001 4:57 PM > Subject: opportunity ? > > > > > > "NASA Explores Future of Software" > > Wall Street Journal (04/26/01) P. A1; Wessel, David > > > > NASA administrator Daniel Goldin is planning to collaborate with > > universities and businesses to develop more reliable > > software at the Ames Research Center in Silicon Valley. Along with > the > > military, NASA has historically led in software > > development for its space projects. As business software matured, > the > > focus of NASA and the corporate sector > > diverged more clearly, but Goldin believes that they may be drawing > near > > again. Henry McDonald, the NASA physicist > > in charge of the Ames center, says that applications need to be > found > > that support both business and NASA > > functions. He says that both large financial firms and NASA need > more > > reliable software to manage reams of digital > > information flowing continuously. Goldin is aware that NASA has > suffered > > software-wise in recent years, and that > > better programs could help prevent disasters like the crash of a > Mars > > Polar Lander in late 1999. NASA has already > > succeeded in signing on one research partner in IBM, who shares the > view > > that such collaborative research can > > improve both the process and the end product. > > > > > ____________________________________________________________________ Get free email and a permanent address at http://www.netaddress.com/?N=1 From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 16:55:13 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id QAA08686 for reliable_computing-outgoing; Sun, 29 Apr 2001 16:55:12 -0500 (CDT) Received: from femail12.sdc1.sfba.home.com (femail12.sdc1.sfba.home.com [24.0.95.108]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id QAA08681 for ; Sun, 29 Apr 2001 16:55:09 -0500 (CDT) Received: from beryl ([65.4.134.233]) by femail12.sdc1.sfba.home.com (InterMail vM.4.01.03.20 201-229-121-120-20010223) with SMTP id <20010429215500.COHR24113.femail12.sdc1.sfba.home.com@beryl>; Sun, 29 Apr 2001 14:55:00 -0700 Message-ID: <00ac01c0d0f8$8ea0bee0$0d0da8c0@beryl> From: "Eldar Musayev \(usa.net\)" To: "Vladik Kreinovich" , References: <200104291935.f3TJZLd22393 [at] cs [dot] utep.edu> Subject: Re: opportunity ? Date: Sun, 29 Apr 2001 15:05:52 -0700 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk True, thanks for clarifying the matter. "One man's problem is another man's opportunity." But it's still a high-hanging fruit. About > From: > From this we conclude that NASA is not interested in avoiding disasters such > as the Hubble Space Telescope?" I have to say two things: (1) I don't work for NASA. (2) I understand that this refers to the problem with an optical aberration distorting one of the lenses, right? Then I fail to see how interval arithmetics can help. Unless you can physically measure and calculate an error based on the result of some measurements, interval error would not depend on a real world and will be useless. If it can be measured, such problems would be detected on a low level. Both ways intervals are out of business. I am more agree with Vladik's example, where intervals are used with the hardware working within the normal parameters. The positive part of my message was about possible connection of interval approach with XP programming techniques, which is going (quite likely) to have a shower of grants, judging from this message. But it's a tough task. ----- Original Message ----- From: "Vladik Kreinovich" To: ; ; Sent: Sunday, April 29, 2001 12:35 PM Subject: Re: opportunity ? > Dear Friends, > > If I may, let me rephrase Eldar's comments in a more positive way. > > Eldar is absolutely right that NASA folks have close to 0 knowledge about > intervals, so if we expect a shower of grants to come our way without us doing > anything, we should not. > > On the other hand, I still believe that because of the emphasis, we have a > window of opportunity to argue that the inaccuracy of input data should be > properly accounted for to make a software really reliable. > > Eldar is absolutely correct that this has not been recognized as a problem in > major NASA applications, but it does not mean it is not a problem: from my side > as a participant in analyzing satellite images from the viewpoint o extracting > geophysical information the inaccuracy of input data is a serious problem. > > Let me also clarify one more reason wht Eldar's pessimistic reaction is not > inconsistent with my optimistic one. (I think we are saying the same thing but > in slightly different terms.) > > Interval computations can take care of two types of uncertainty: > * uncertainty caused by rounding errors, and > * interval uncertainty of the input data (coming from measurements with unknown > error probability distribution but known upper bounds on the measurement > error). > > In numerical mathematics, errors of the first type are often the main problem, > which explains the widely spread emphasis on interval computations as a better > alternative to floating-point arithmetic. I agree with Eldar 100% that this is > not the main source of NASA problems ("NASA problems are usually not with the > floating-point arithmetic"). > > In NASA applications, the main source of errors is the errors of the second > type, and here, the situation when we do not know the error probability > distribution is even more widely spread than in other areas. > > After all, to get a probability distribution for errors, we usually compare the > measurement results with results obtained by using some more accuracte > (standard) measuring instrument. However, for Hubble telescope (and, for that > reason, for a satellite which sends us photos of Earth), there is no standard > telescope floating nearby which can be used to calibrate the original > instrument. At best, we get upper bounds on the measurement errors, and when > you have upper bounds D on the error, every measurement results in an interval > [x-D,x+D] of possible values. > > When we process the data, we transform the measured values x1,...,xn into some > value y=f(x1,...,xn), for some function f given usually implicitly by a > program. If we only know intervals for the inputs, then the problem of > computing the interval for the output becomes a typical problem of interval > computations. > > Do NASA people know it? Most don't, Eldar is right about it. However, with the > emphasis on reliability of software, we have a better chance to get their > attention and their ears. This is why I view it as an opportunity to succeed > (but not a guarantee that we will succeed). > > When a few years ago, Dan Goldin visited El Paso and our NASA-sponsored center, > I was impressed by the fact that he did realize the necessity to use > intelligent methods. I know that this realization is the result of a hard > education work, persistence, persistence, and persistence. I did give my spiel > about intervals then, but of course, it is not enough. > > We need to use every effort to educate NASA folks and maybe, they will start > understanding the need for interval computations as well. The best way to > educate them is to show that it works on their real-life problems. I think > control people have a good chance of using interval-based robust control, etc. > With the new worry, we have a slightly better chance to get our foot in. > > Vladik > > > From: "Eldar Musayev \(usa.net\)" > > To: "Vladik Kreinovich" , > > > Subject: Re: opportunity ? > > Date: Sun, 29 Apr 2001 12:16:40 -0700 > > MIME-Version: 1.0 > > Content-Transfer-Encoding: 7bit > > X-Priority: 3 > > X-MSMail-Priority: Normal > > X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 > > > > Generally, no. > > > > NASA problems are usually not with the floating-point arithmetic, but with > > generic software bugs, which call for a new level of theoritization in > > Computer Science about how to write reliable (no generic bugs, not a > > computation per se) programs, how to "control those stupid programmers > > without doing the job for them" (which is, luckily for researchers, anyway > > impossible both ways), with grants for researchers in these areas, with new > > Rational (and probably Irrational too) Roses etc. And yes, it's a lot of > > sarcasm about such techniques, though not all of them. I think that as a > > direct result, you may expect a burst of publications on XP programming, so > > if you can link your research with XP programming, it may be a reasonable > > shot. There could be some other prospective areas, but they seems to be even > > farther from interval arithmetic. So otherwise, chances seems to be dim. > > > > -- Eldar > > > > > > ----- Original Message ----- > > From: "Vladik Kreinovich" > > To: > > Sent: Friday, April 27, 2001 4:57 PM > > Subject: opportunity ? > > > > > > > > > > "NASA Explores Future of Software" > > > Wall Street Journal (04/26/01) P. A1; Wessel, David > > > > > > NASA administrator Daniel Goldin is planning to collaborate with > > > universities and businesses to develop more reliable > > > software at the Ames Research Center in Silicon Valley. Along with > > the > > > military, NASA has historically led in software > > > development for its space projects. As business software matured, > > the > > > focus of NASA and the corporate sector > > > diverged more clearly, but Goldin believes that they may be drawing > > near > > > again. Henry McDonald, the NASA physicist > > > in charge of the Ames center, says that applications need to be > > found > > > that support both business and NASA > > > functions. He says that both large financial firms and NASA need > > more > > > reliable software to manage reams of digital > > > information flowing continuously. Goldin is aware that NASA has > > suffered > > > software-wise in recent years, and that > > > better programs could help prevent disasters like the crash of a > > Mars > > > Polar Lander in late 1999. NASA has already > > > succeeded in signing on one research partner in IBM, who shares the > > view > > > that such collaborative research can > > > improve both the process and the end product. > > > > > > > > From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 18:24:19 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id SAA09041 for reliable_computing-outgoing; Sun, 29 Apr 2001 18:24:18 -0500 (CDT) Received: from nwcst320.netaddress.usa.net (nwcst320.netaddress.usa.net [204.68.23.65]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with SMTP id SAA09036 for ; Sun, 29 Apr 2001 18:24:15 -0500 (CDT) From: ljiw [at] usa [dot] net Received: (qmail 13916 invoked by uid 60001); 29 Apr 2001 23:24:13 -0000 Message-ID: <20010429232413.13915.qmail [at] nwcst320 [dot] netaddress.usa.net> Received: from 204.68.23.65 by nwcst320 for [24.128.176.15] via web-mailer(34FM.0700.17C.01) on Sun Apr 29 23:24:13 GMT 2001 Date: 29 Apr 2001 19:24:13 EDT To: "Eldar Musayev \(usa.net\)" , "Vladik Kreinovich" , Subject: Re: [Re: opportunity ?] X-Mailer: USANET web-mailer (34FM.0700.17C.01) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by interval.usl.edu id SAA09037 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk "Eldar Musayev \(usa.net\)" wrote: > True, thanks for clarifying the matter. "One man's problem is another man's > opportunity." > But it's still a high-hanging fruit. > > About > > From: > > From this we conclude that NASA is not interested in avoiding disasters > such > > as the Hubble Space Telescope?" > I have to say two things: > (1) I don't work for NASA. > (2) I understand that this refers to the problem with an optical aberration > distorting one of the lenses, right? Not quite. As I understand it there was a very thorough testing protocol used in the construction of the lens. It was measured against a refernce surface that was designed for that particular measurement. But the design of the reference was performed on a spreadsheet and no numerical analysis of the model was available. It turns out that the model had an important loss of precision that affected the shape of the reference surface. So the lens was properly built, but to improper specs. And the spec were improper due to a loss of precision that the use of interval arithmetic would have made obvious. > Then I fail to see how interval arithmetics can help. Unless you can > physically measure and calculate > an error based on the result of some measurements, interval error would not > depend on a real world and will > be useless. If it can be measured, such problems would be detected on a low > level. Both ways intervals are out of business. > I am more agree with Vladik's example, where intervals are used with the > hardware working within the normal parameters. > > The positive part of my message was about possible connection of interval > approach with XP programming techniques, > which is going (quite likely) to have a shower of grants, judging from this > message. But it's a tough task. > > > ----- Original Message ----- > From: "Vladik Kreinovich" > To: ; ; > > Sent: Sunday, April 29, 2001 12:35 PM > Subject: Re: opportunity ? > > > > Dear Friends, > > > > If I may, let me rephrase Eldar's comments in a more positive way. > > > > Eldar is absolutely right that NASA folks have close to 0 knowledge about > > intervals, so if we expect a shower of grants to come our way without us > doing > > anything, we should not. > > > > On the other hand, I still believe that because of the emphasis, we have a > > window of opportunity to argue that the inaccuracy of input data should be > > properly accounted for to make a software really reliable. > > > > Eldar is absolutely correct that this has not been recognized as a problem > in > > major NASA applications, but it does not mean it is not a problem: from my > side > > as a participant in analyzing satellite images from the viewpoint o > extracting > > geophysical information the inaccuracy of input data is a serious problem. > > > > Let me also clarify one more reason wht Eldar's pessimistic reaction is > not > > inconsistent with my optimistic one. (I think we are saying the same thing > but > > in slightly different terms.) > > > > Interval computations can take care of two types of uncertainty: > > * uncertainty caused by rounding errors, and > > * interval uncertainty of the input data (coming from measurements with > unknown > > error probability distribution but known upper bounds on the measurement > > error). > > > > In numerical mathematics, errors of the first type are often the main > problem, > > which explains the widely spread emphasis on interval computations as a > better > > alternative to floating-point arithmetic. I agree with Eldar 100% that > this is > > not the main source of NASA problems ("NASA problems are usually not with > the > > floating-point arithmetic"). > > > > In NASA applications, the main source of errors is the errors of the > second > > type, and here, the situation when we do not know the error probability > > distribution is even more widely spread than in other areas. > > > > After all, to get a probability distribution for errors, we usually > compare the > > measurement results with results obtained by using some more accuracte > > (standard) measuring instrument. However, for Hubble telescope (and, for > that > > reason, for a satellite which sends us photos of Earth), there is no > standard > > telescope floating nearby which can be used to calibrate the original > > instrument. At best, we get upper bounds on the measurement errors, and > when > > you have upper bounds D on the error, every measurement results in an > interval > > [x-D,x+D] of possible values. > > > > When we process the data, we transform the measured values x1,...,xn into > some > > value y=f(x1,...,xn), for some function f given usually implicitly by a > > program. If we only know intervals for the inputs, then the problem of > > computing the interval for the output becomes a typical problem of > interval > > computations. > > > > Do NASA people know it? Most don't, Eldar is right about it. However, with > the > > emphasis on reliability of software, we have a better chance to get their > > attention and their ears. This is why I view it as an opportunity to > succeed > > (but not a guarantee that we will succeed). > > > > When a few years ago, Dan Goldin visited El Paso and our NASA-sponsored > center, > > I was impressed by the fact that he did realize the necessity to use > > intelligent methods. I know that this realization is the result of a hard > > education work, persistence, persistence, and persistence. I did give my > spiel > > about intervals then, but of course, it is not enough. > > > > We need to use every effort to educate NASA folks and maybe, they will > start > > understanding the need for interval computations as well. The best way to > > educate them is to show that it works on their real-life problems. I think > > control people have a good chance of using interval-based robust control, > etc. > > With the new worry, we have a slightly better chance to get our foot in. > > > > Vladik > > > > > From: "Eldar Musayev \(usa.net\)" > > > To: "Vladik Kreinovich" , > > > > > Subject: Re: opportunity ? > > > Date: Sun, 29 Apr 2001 12:16:40 -0700 > > > MIME-Version: 1.0 > > > Content-Transfer-Encoding: 7bit > > > X-Priority: 3 > > > X-MSMail-Priority: Normal > > > X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 > > > > > > Generally, no. > > > > > > NASA problems are usually not with the floating-point arithmetic, but > with > > > generic software bugs, which call for a new level of theoritization in > > > Computer Science about how to write reliable (no generic bugs, not a > > > computation per se) programs, how to "control those stupid programmers > > > without doing the job for them" (which is, luckily for researchers, > anyway > > > impossible both ways), with grants for researchers in these areas, with > new > > > Rational (and probably Irrational too) Roses etc. And yes, it's a lot of > > > sarcasm about such techniques, though not all of them. I think that as a > > > direct result, you may expect a burst of publications on XP programming, > so > > > if you can link your research with XP programming, it may be a > reasonable > > > shot. There could be some other prospective areas, but they seems to be > even > > > farther from interval arithmetic. So otherwise, chances seems to be dim. > > > > > > -- Eldar > > > > > > > > > ----- Original Message ----- > > > From: "Vladik Kreinovich" > > > To: > > > Sent: Friday, April 27, 2001 4:57 PM > > > Subject: opportunity ? > > > > > > > > > > > > > > "NASA Explores Future of Software" > > > > Wall Street Journal (04/26/01) P. A1; Wessel, David > > > > > > > > NASA administrator Daniel Goldin is planning to collaborate with > > > > universities and businesses to develop more reliable > > > > software at the Ames Research Center in Silicon Valley. Along > with > > > the > > > > military, NASA has historically led in software > > > > development for its space projects. As business software > matured, > > > the > > > > focus of NASA and the corporate sector > > > > diverged more clearly, but Goldin believes that they may be > drawing > > > near > > > > again. Henry McDonald, the NASA physicist > > > > in charge of the Ames center, says that applications need to be > > > found > > > > that support both business and NASA > > > > functions. He says that both large financial firms and NASA need > > > more > > > > reliable software to manage reams of digital > > > > information flowing continuously. Goldin is aware that NASA has > > > suffered > > > > software-wise in recent years, and that > > > > better programs could help prevent disasters like the crash of a > > > Mars > > > > Polar Lander in late 1999. NASA has already > > > > succeeded in signing on one research partner in IBM, who shares > the > > > view > > > > that such collaborative research can > > > > improve both the process and the end product. > > > > > > > > > > > > > ____________________________________________________________________ Get free email and a permanent address at http://www.netaddress.com/?N=1 From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 19:26:23 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id TAA09369 for reliable_computing-outgoing; Sun, 29 Apr 2001 19:26:23 -0500 (CDT) Received: from femail12.sdc1.sfba.home.com (femail12.sdc1.sfba.home.com [24.0.95.108]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id TAA09364 for ; Sun, 29 Apr 2001 19:26:19 -0500 (CDT) Received: from beryl ([65.4.134.233]) by femail12.sdc1.sfba.home.com (InterMail vM.4.01.03.20 201-229-121-120-20010223) with SMTP id <20010430002613.GJKB24113.femail12.sdc1.sfba.home.com@beryl> for ; Sun, 29 Apr 2001 17:26:13 -0700 Message-ID: <00f001c0d10d$aeab3610$0d0da8c0@beryl> From: "Eldar Musayev \(usa.net\)" To: References: <20010429232413.13915.qmail [at] nwcst320 [dot] netaddress.usa.net> Subject: Re: [Re: opportunity ?] Date: Sun, 29 Apr 2001 17:37:05 -0700 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Seems like a very good selling point. Though, if they used a spreadsheet, I doubt they would use a program, either interval or not. Anyway, it's a good story for interval sites, to show IA importance, the ones like Vladik or Baker maintain. ----- Original Message ----- From: To: "Eldar Musayev (usa.net)" ; "Vladik Kreinovich" ; Sent: Sunday, April 29, 2001 4:24 PM Subject: Re: [Re: opportunity ?] > "Eldar Musayev \(usa.net\)" wrote: > > True, thanks for clarifying the matter. "One man's problem is another man's > > opportunity." > > But it's still a high-hanging fruit. > > > > About > > > From: > > > From this we conclude that NASA is not interested in avoiding disasters > > such > > > as the Hubble Space Telescope?" > > I have to say two things: > > (1) I don't work for NASA. > > (2) I understand that this refers to the problem with an optical aberration > > distorting one of the lenses, right? > > Not quite. As I understand it there was a very thorough testing protocol used > in the construction of the lens. It was measured against a refernce surface > that was designed for that particular measurement. But the design of the > reference was performed on a spreadsheet and no numerical analysis of the > model was available. It turns out that the model had an important loss of > precision that affected the shape of the reference surface. > > So the lens was properly built, but to improper specs. And the spec were > improper due to a loss of precision that the use of interval arithmetic would > have made obvious. > > > Then I fail to see how interval arithmetics can help. Unless you can > > physically measure and calculate > > an error based on the result of some measurements, interval error would not > > depend on a real world and will > > be useless. If it can be measured, such problems would be detected on a low > > level. Both ways intervals are out of business. > > I am more agree with Vladik's example, where intervals are used with the > > hardware working within the normal parameters. > > > > The positive part of my message was about possible connection of interval > > approach with XP programming techniques, > > which is going (quite likely) to have a shower of grants, judging from this > > message. But it's a tough task. > > > > > > ----- Original Message ----- > > From: "Vladik Kreinovich" > > To: ; ; > > > > Sent: Sunday, April 29, 2001 12:35 PM > > Subject: Re: opportunity ? > > > > > > > Dear Friends, > > > > > > If I may, let me rephrase Eldar's comments in a more positive way. > > > > > > Eldar is absolutely right that NASA folks have close to 0 knowledge about > > > intervals, so if we expect a shower of grants to come our way without us > > doing > > > anything, we should not. > > > > > > On the other hand, I still believe that because of the emphasis, we have > a > > > window of opportunity to argue that the inaccuracy of input data should > be > > > properly accounted for to make a software really reliable. > > > > > > Eldar is absolutely correct that this has not been recognized as a > problem > > in > > > major NASA applications, but it does not mean it is not a problem: from > my > > side > > > as a participant in analyzing satellite images from the viewpoint o > > extracting > > > geophysical information the inaccuracy of input data is a serious > problem. > > > > > > Let me also clarify one more reason wht Eldar's pessimistic reaction is > > not > > > inconsistent with my optimistic one. (I think we are saying the same > thing > > but > > > in slightly different terms.) > > > > > > Interval computations can take care of two types of uncertainty: > > > * uncertainty caused by rounding errors, and > > > * interval uncertainty of the input data (coming from measurements with > > unknown > > > error probability distribution but known upper bounds on the measurement > > > error). > > > > > > In numerical mathematics, errors of the first type are often the main > > problem, > > > which explains the widely spread emphasis on interval computations as a > > better > > > alternative to floating-point arithmetic. I agree with Eldar 100% that > > this is > > > not the main source of NASA problems ("NASA problems are usually not with > > the > > > floating-point arithmetic"). > > > > > > In NASA applications, the main source of errors is the errors of the > > second > > > type, and here, the situation when we do not know the error probability > > > distribution is even more widely spread than in other areas. > > > > > > After all, to get a probability distribution for errors, we usually > > compare the > > > measurement results with results obtained by using some more accuracte > > > (standard) measuring instrument. However, for Hubble telescope (and, for > > that > > > reason, for a satellite which sends us photos of Earth), there is no > > standard > > > telescope floating nearby which can be used to calibrate the original > > > instrument. At best, we get upper bounds on the measurement errors, and > > when > > > you have upper bounds D on the error, every measurement results in an > > interval > > > [x-D,x+D] of possible values. > > > > > > When we process the data, we transform the measured values x1,...,xn into > > some > > > value y=f(x1,...,xn), for some function f given usually implicitly by a > > > program. If we only know intervals for the inputs, then the problem of > > > computing the interval for the output becomes a typical problem of > > interval > > > computations. > > > > > > Do NASA people know it? Most don't, Eldar is right about it. However, > with > > the > > > emphasis on reliability of software, we have a better chance to get their > > > attention and their ears. This is why I view it as an opportunity to > > succeed > > > (but not a guarantee that we will succeed). > > > > > > When a few years ago, Dan Goldin visited El Paso and our NASA-sponsored > > center, > > > I was impressed by the fact that he did realize the necessity to use > > > intelligent methods. I know that this realization is the result of a hard > > > education work, persistence, persistence, and persistence. I did give my > > spiel > > > about intervals then, but of course, it is not enough. > > > > > > We need to use every effort to educate NASA folks and maybe, they will > > start > > > understanding the need for interval computations as well. The best way to > > > educate them is to show that it works on their real-life problems. I > think > > > control people have a good chance of using interval-based robust control, > > etc. > > > With the new worry, we have a slightly better chance to get our foot in. > > > > > > Vladik > > > > > > > From: "Eldar Musayev \(usa.net\)" > > > > To: "Vladik Kreinovich" , > > > > > > > Subject: Re: opportunity ? > > > > Date: Sun, 29 Apr 2001 12:16:40 -0700 > > > > MIME-Version: 1.0 > > > > Content-Transfer-Encoding: 7bit > > > > X-Priority: 3 > > > > X-MSMail-Priority: Normal > > > > X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 > > > > > > > > Generally, no. > > > > > > > > NASA problems are usually not with the floating-point arithmetic, but > > with > > > > generic software bugs, which call for a new level of theoritization in > > > > Computer Science about how to write reliable (no generic bugs, not a > > > > computation per se) programs, how to "control those stupid programmers > > > > without doing the job for them" (which is, luckily for researchers, > > anyway > > > > impossible both ways), with grants for researchers in these areas, with > > new > > > > Rational (and probably Irrational too) Roses etc. And yes, it's a lot > of > > > > sarcasm about such techniques, though not all of them. I think that as > a > > > > direct result, you may expect a burst of publications on XP > programming, > > so > > > > if you can link your research with XP programming, it may be a > > reasonable > > > > shot. There could be some other prospective areas, but they seems to be > > even > > > > farther from interval arithmetic. So otherwise, chances seems to be > dim. > > > > > > > > -- Eldar > > > > > > > > > > > > ----- Original Message ----- > > > > From: "Vladik Kreinovich" > > > > To: > > > > Sent: Friday, April 27, 2001 4:57 PM > > > > Subject: opportunity ? > > > > > > > > > > > > > > > > > > "NASA Explores Future of Software" > > > > > Wall Street Journal (04/26/01) P. A1; Wessel, David > > > > > > > > > > NASA administrator Daniel Goldin is planning to collaborate > with > > > > > universities and businesses to develop more reliable > > > > > software at the Ames Research Center in Silicon Valley. Along > > with > > > > the > > > > > military, NASA has historically led in software > > > > > development for its space projects. As business software > > matured, > > > > the > > > > > focus of NASA and the corporate sector > > > > > diverged more clearly, but Goldin believes that they may be > > drawing > > > > near > > > > > again. Henry McDonald, the NASA physicist > > > > > in charge of the Ames center, says that applications need to be > > > > found > > > > > that support both business and NASA > > > > > functions. He says that both large financial firms and NASA > need > > > > more > > > > > reliable software to manage reams of digital > > > > > information flowing continuously. Goldin is aware that NASA has > > > > suffered > > > > > software-wise in recent years, and that > > > > > better programs could help prevent disasters like the crash of > a > > > > Mars > > > > > Polar Lander in late 1999. NASA has already > > > > > succeeded in signing on one research partner in IBM, who shares > > the > > > > view > > > > > that such collaborative research can > > > > > improve both the process and the end product. > > > > > > > > > > > > > > > > > > > > > ____________________________________________________________________ > Get free email and a permanent address at http://www.netaddress.com/?N=1 > From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 19:57:40 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id TAA09724 for reliable_computing-outgoing; Sun, 29 Apr 2001 19:57:40 -0500 (CDT) Received: from cse.nd.edu (root [at] wizard [dot] cse.nd.edu [129.74.25.101]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id TAA09719 for ; Sun, 29 Apr 2001 19:57:36 -0500 (CDT) Received: from Tyler (root@localhost [127.0.0.1]) (authenticated) by cse.nd.edu (8.11.2/8.11.2) with ESMTP id f3U0vKo21364; Sun, 29 Apr 2001 19:57:20 -0500 (EST) Message-ID: <007801c0d110$5dc000c0$5d63a8c0@Tyler> From: "Andrew Lumsdaine" To: , "Eldar Musayev \(usa.net\)" , "Vladik Kreinovich" , References: <20010429232413.13915.qmail [at] nwcst320 [dot] netaddress.usa.net> Subject: Re: [Re: opportunity ?] Date: Sun, 29 Apr 2001 19:56:13 -0500 Organization: University of Notre Dame MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4133.2400 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Er, a couple of comments -- this is way off-topic, but I can't resist. First, the problem with the Hubble was in the primary mirror, not a lens. Second, the problems in manufacturing seem to be due primarily to human error, not problems with numerical computation. See, e.g., page 511-512 in http://history.msfc.nasa.gov/book/chpttwelve.pdf (the whole thing is quite an interesting read). Maybe it is just because I am not an expert in intervals, but I fail to understand how interval computations could have prevented this particular problem. Slightly more on topic. There is a fundamental (but perhaps subtle) difference between reliable software and reliable (numerical) computing. As has already been pointed out, NASA is interested in attempting to stamp out software problems of all kinds, not just those due to numerical imprecision. The amount of software written for numerical purposes is surprisingly small. ----- Original Message ----- From: To: "Eldar Musayev (usa.net)" ; "Vladik Kreinovich" ; Sent: Sunday, April 29, 2001 6:24 PM Subject: Re: [Re: opportunity ?] > "Eldar Musayev \(usa.net\)" wrote: > > True, thanks for clarifying the matter. "One man's problem is another man's > > opportunity." > > But it's still a high-hanging fruit. > > > > About > > > From: > > > From this we conclude that NASA is not interested in avoiding disasters > > such > > > as the Hubble Space Telescope?" > > I have to say two things: > > (1) I don't work for NASA. > > (2) I understand that this refers to the problem with an optical aberration > > distorting one of the lenses, right? > > Not quite. As I understand it there was a very thorough testing protocol used > in the construction of the lens. It was measured against a refernce surface > that was designed for that particular measurement. But the design of the > reference was performed on a spreadsheet and no numerical analysis of the > model was available. It turns out that the model had an important loss of > precision that affected the shape of the reference surface. > > So the lens was properly built, but to improper specs. And the spec were > improper due to a loss of precision that the use of interval arithmetic would > have made obvious. > > > Then I fail to see how interval arithmetics can help. Unless you can > > physically measure and calculate > > an error based on the result of some measurements, interval error would not > > depend on a real world and will > > be useless. If it can be measured, such problems would be detected on a low > > level. Both ways intervals are out of business. > > I am more agree with Vladik's example, where intervals are used with the > > hardware working within the normal parameters. > > > > The positive part of my message was about possible connection of interval > > approach with XP programming techniques, > > which is going (quite likely) to have a shower of grants, judging from this > > message. But it's a tough task. > > > > > > ----- Original Message ----- > > From: "Vladik Kreinovich" > > To: ; ; > > > > Sent: Sunday, April 29, 2001 12:35 PM > > Subject: Re: opportunity ? > > > > > > > Dear Friends, > > > > > > If I may, let me rephrase Eldar's comments in a more positive way. > > > > > > Eldar is absolutely right that NASA folks have close to 0 knowledge about > > > intervals, so if we expect a shower of grants to come our way without us > > doing > > > anything, we should not. > > > > > > On the other hand, I still believe that because of the emphasis, we have > a > > > window of opportunity to argue that the inaccuracy of input data should > be > > > properly accounted for to make a software really reliable. > > > > > > Eldar is absolutely correct that this has not been recognized as a > problem > > in > > > major NASA applications, but it does not mean it is not a problem: from > my > > side > > > as a participant in analyzing satellite images from the viewpoint o > > extracting > > > geophysical information the inaccuracy of input data is a serious > problem. > > > > > > Let me also clarify one more reason wht Eldar's pessimistic reaction is > > not > > > inconsistent with my optimistic one. (I think we are saying the same > thing > > but > > > in slightly different terms.) > > > > > > Interval computations can take care of two types of uncertainty: > > > * uncertainty caused by rounding errors, and > > > * interval uncertainty of the input data (coming from measurements with > > unknown > > > error probability distribution but known upper bounds on the measurement > > > error). > > > > > > In numerical mathematics, errors of the first type are often the main > > problem, > > > which explains the widely spread emphasis on interval computations as a > > better > > > alternative to floating-point arithmetic. I agree with Eldar 100% that > > this is > > > not the main source of NASA problems ("NASA problems are usually not with > > the > > > floating-point arithmetic"). > > > > > > In NASA applications, the main source of errors is the errors of the > > second > > > type, and here, the situation when we do not know the error probability > > > distribution is even more widely spread than in other areas. > > > > > > After all, to get a probability distribution for errors, we usually > > compare the > > > measurement results with results obtained by using some more accuracte > > > (standard) measuring instrument. However, for Hubble telescope (and, for > > that > > > reason, for a satellite which sends us photos of Earth), there is no > > standard > > > telescope floating nearby which can be used to calibrate the original > > > instrument. At best, we get upper bounds on the measurement errors, and > > when > > > you have upper bounds D on the error, every measurement results in an > > interval > > > [x-D,x+D] of possible values. > > > > > > When we process the data, we transform the measured values x1,...,xn into > > some > > > value y=f(x1,...,xn), for some function f given usually implicitly by a > > > program. If we only know intervals for the inputs, then the problem of > > > computing the interval for the output becomes a typical problem of > > interval > > > computations. > > > > > > Do NASA people know it? Most don't, Eldar is right about it. However, > with > > the > > > emphasis on reliability of software, we have a better chance to get their > > > attention and their ears. This is why I view it as an opportunity to > > succeed > > > (but not a guarantee that we will succeed). > > > > > > When a few years ago, Dan Goldin visited El Paso and our NASA-sponsored > > center, > > > I was impressed by the fact that he did realize the necessity to use > > > intelligent methods. I know that this realization is the result of a hard > > > education work, persistence, persistence, and persistence. I did give my > > spiel > > > about intervals then, but of course, it is not enough. > > > > > > We need to use every effort to educate NASA folks and maybe, they will > > start > > > understanding the need for interval computations as well. The best way to > > > educate them is to show that it works on their real-life problems. I > think > > > control people have a good chance of using interval-based robust control, > > etc. > > > With the new worry, we have a slightly better chance to get our foot in. > > > > > > Vladik > > > > > > > From: "Eldar Musayev \(usa.net\)" > > > > To: "Vladik Kreinovich" , > > > > > > > Subject: Re: opportunity ? > > > > Date: Sun, 29 Apr 2001 12:16:40 -0700 > > > > MIME-Version: 1.0 > > > > Content-Transfer-Encoding: 7bit > > > > X-Priority: 3 > > > > X-MSMail-Priority: Normal > > > > X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 > > > > > > > > Generally, no. > > > > > > > > NASA problems are usually not with the floating-point arithmetic, but > > with > > > > generic software bugs, which call for a new level of theoritization in > > > > Computer Science about how to write reliable (no generic bugs, not a > > > > computation per se) programs, how to "control those stupid programmers > > > > without doing the job for them" (which is, luckily for researchers, > > anyway > > > > impossible both ways), with grants for researchers in these areas, with > > new > > > > Rational (and probably Irrational too) Roses etc. And yes, it's a lot > of > > > > sarcasm about such techniques, though not all of them. I think that as > a > > > > direct result, you may expect a burst of publications on XP > programming, > > so > > > > if you can link your research with XP programming, it may be a > > reasonable > > > > shot. There could be some other prospective areas, but they seems to be > > even > > > > farther from interval arithmetic. So otherwise, chances seems to be > dim. > > > > > > > > -- Eldar > > > > > > > > > > > > ----- Original Message ----- > > > > From: "Vladik Kreinovich" > > > > To: > > > > Sent: Friday, April 27, 2001 4:57 PM > > > > Subject: opportunity ? > > > > > > > > > > > > > > > > > > "NASA Explores Future of Software" > > > > > Wall Street Journal (04/26/01) P. A1; Wessel, David > > > > > > > > > > NASA administrator Daniel Goldin is planning to collaborate > with > > > > > universities and businesses to develop more reliable > > > > > software at the Ames Research Center in Silicon Valley. Along > > with > > > > the > > > > > military, NASA has historically led in software > > > > > development for its space projects. As business software > > matured, > > > > the > > > > > focus of NASA and the corporate sector > > > > > diverged more clearly, but Goldin believes that they may be > > drawing > > > > near > > > > > again. Henry McDonald, the NASA physicist > > > > > in charge of the Ames center, says that applications need to be > > > > found > > > > > that support both business and NASA > > > > > functions. He says that both large financial firms and NASA > need > > > > more > > > > > reliable software to manage reams of digital > > > > > information flowing continuously. Goldin is aware that NASA has > > > > suffered > > > > > software-wise in recent years, and that > > > > > better programs could help prevent disasters like the crash of > a > > > > Mars > > > > > Polar Lander in late 1999. NASA has already > > > > > succeeded in signing on one research partner in IBM, who shares > > the > > > > view > > > > > that such collaborative research can > > > > > improve both the process and the end product. > > > > > > > > > > > > > > > > > > > > > ____________________________________________________________________ > Get free email and a permanent address at http://www.netaddress.com/?N=1 > From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 20:53:50 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id UAA10049 for reliable_computing-outgoing; Sun, 29 Apr 2001 20:53:50 -0500 (CDT) Received: from nw175.netaddress.usa.net (nw175.netaddress.usa.net [204.68.24.75]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with SMTP id UAA10044 for ; Sun, 29 Apr 2001 20:53:46 -0500 (CDT) From: ljiw [at] usa [dot] net Received: (qmail 7845 invoked by uid 60001); 30 Apr 2001 01:53:43 -0000 Message-ID: <20010430015343.7844.qmail [at] nw175 [dot] netaddress.usa.net> Received: from 204.68.24.75 by nw175 for [24.128.176.15] via web-mailer(34FM.0700.17C.01) on Mon Apr 30 01:53:43 GMT 2001 Date: 29 Apr 2001 21:53:43 EDT To: "Andrew Lumsdaine" , , "Eldar Musayev \(usa.net\)" , "Vladik Kreinovich" , Subject: Re: [Re: [Re: opportunity ?]] X-Mailer: USANET web-mailer (34FM.0700.17C.01) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by interval.usl.edu id UAA10045 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk "Andrew Lumsdaine" wrote: > Er, a couple of comments -- this is way off-topic, but I can't resist. > First, the problem with the Hubble was in the primary mirror, not a lens. Yes, the main first-surface reflector. I typed too quickly. > Second, the problems in manufacturing seem to be due primarily to human > error, not problems with numerical computation. See, e.g., page 511-512 in > http://history.msfc.nasa.gov/book/chpttwelve.pdf (the whole thing is quite > an interesting read). The chapter referenced abstracts an explanation from a NASA report (the Allen report) that mentions misreading a scale by 1.3 mm. But the reading I did around the time of the event (which I'll have to dig up) indicated that the reference surface (AKA the null corrector) was improperly designed. Maybe it is just because I am not an expert in > intervals, but I fail to understand how interval computations could have > prevented this particular problem. If the problem was a technician misreading a measuring device no software could influence the outcome. > > Slightly more on topic. There is a fundamental (but perhaps subtle) > difference between reliable software and reliable (numerical) computing. As > has already been pointed out, NASA is interested in attempting to stamp out > software problems of all kinds, not just those due to numerical imprecision. > The amount of software written for numerical purposes is surprisingly small. And the amount of software that needs numerical validity is far larger than that written for numerical purposes. E.g., analyzing an electrical circuit requires an evaluation of the tolerances for each component. Yet such software routinely uses the nominal value, about which there is a 1%, 5%, or 10% error region, as the actual value and computes the voltage and/or current with no consideration for the possible variations. This is hardly a numerical purpose. Yet it is a perfectly suitable applicaiton of interval arithmetic. > > ----- Original Message ----- > From: > To: "Eldar Musayev (usa.net)" ; "Vladik Kreinovich" > ; > Sent: Sunday, April 29, 2001 6:24 PM > Subject: Re: [Re: opportunity ?] > > > > "Eldar Musayev \(usa.net\)" wrote: > > > True, thanks for clarifying the matter. "One man's problem is another > man's > > > opportunity." > > > But it's still a high-hanging fruit. > > > > > > About > > > > From: > > > > From this we conclude that NASA is not interested in avoiding > disasters > > > such > > > > as the Hubble Space Telescope?" > > > I have to say two things: > > > (1) I don't work for NASA. > > > (2) I understand that this refers to the problem with an optical > aberration > > > distorting one of the lenses, right? > > > > Not quite. As I understand it there was a very thorough testing protocol > used > > in the construction of the lens. It was measured against a refernce > surface > > that was designed for that particular measurement. But the design of the > > reference was performed on a spreadsheet and no numerical analysis of the > > model was available. It turns out that the model had an important loss of > > precision that affected the shape of the reference surface. > > > > So the lens was properly built, but to improper specs. And the spec were > > improper due to a loss of precision that the use of interval arithmetic > would > > have made obvious. > > > > > Then I fail to see how interval arithmetics can help. Unless you can > > > physically measure and calculate > > > an error based on the result of some measurements, interval error would > not > > > depend on a real world and will > > > be useless. If it can be measured, such problems would be detected on a > low > > > level. Both ways intervals are out of business. > > > I am more agree with Vladik's example, where intervals are used with the > > > hardware working within the normal parameters. > > > > > > The positive part of my message was about possible connection of > interval > > > approach with XP programming techniques, > > > which is going (quite likely) to have a shower of grants, judging from > this > > > message. But it's a tough task. > > > > > > > > > ----- Original Message ----- > > > From: "Vladik Kreinovich" > > > To: ; ; > > > > > > Sent: Sunday, April 29, 2001 12:35 PM > > > Subject: Re: opportunity ? > > > > > > > > > > Dear Friends, > > > > > > > > If I may, let me rephrase Eldar's comments in a more positive way. > > > > > > > > Eldar is absolutely right that NASA folks have close to 0 knowledge > about > > > > intervals, so if we expect a shower of grants to come our way without > us > > > doing > > > > anything, we should not. > > > > > > > > On the other hand, I still believe that because of the emphasis, we > have > > a > > > > window of opportunity to argue that the inaccuracy of input data > should > > be > > > > properly accounted for to make a software really reliable. > > > > > > > > Eldar is absolutely correct that this has not been recognized as a > > problem > > > in > > > > major NASA applications, but it does not mean it is not a problem: > from > > my > > > side > > > > as a participant in analyzing satellite images from the viewpoint o > > > extracting > > > > geophysical information the inaccuracy of input data is a serious > > problem. > > > > > > > > Let me also clarify one more reason wht Eldar's pessimistic reaction > is > > > not > > > > inconsistent with my optimistic one. (I think we are saying the same > > thing > > > but > > > > in slightly different terms.) > > > > > > > > Interval computations can take care of two types of uncertainty: > > > > * uncertainty caused by rounding errors, and > > > > * interval uncertainty of the input data (coming from measurements > with > > > unknown > > > > error probability distribution but known upper bounds on the > measurement > > > > error). > > > > > > > > In numerical mathematics, errors of the first type are often the main > > > problem, > > > > which explains the widely spread emphasis on interval computations as > a > > > better > > > > alternative to floating-point arithmetic. I agree with Eldar 100% that > > > this is > > > > not the main source of NASA problems ("NASA problems are usually not > with > > > the > > > > floating-point arithmetic"). > > > > > > > > In NASA applications, the main source of errors is the errors of the > > > second > > > > type, and here, the situation when we do not know the error > probability > > > > distribution is even more widely spread than in other areas. > > > > > > > > After all, to get a probability distribution for errors, we usually > > > compare the > > > > measurement results with results obtained by using some more accuracte > > > > (standard) measuring instrument. However, for Hubble telescope (and, > for > > > that > > > > reason, for a satellite which sends us photos of Earth), there is no > > > standard > > > > telescope floating nearby which can be used to calibrate the original > > > > instrument. At best, we get upper bounds on the measurement errors, > and > > > when > > > > you have upper bounds D on the error, every measurement results in an > > > interval > > > > [x-D,x+D] of possible values. > > > > > > > > When we process the data, we transform the measured values x1,...,xn > into > > > some > > > > value y=f(x1,...,xn), for some function f given usually implicitly by > a > > > > program. If we only know intervals for the inputs, then the problem of > > > > computing the interval for the output becomes a typical problem of > > > interval > > > > computations. > > > > > > > > Do NASA people know it? Most don't, Eldar is right about it. However, > > with > > > the > > > > emphasis on reliability of software, we have a better chance to get > their > > > > attention and their ears. This is why I view it as an opportunity to > > > succeed > > > > (but not a guarantee that we will succeed). > > > > > > > > When a few years ago, Dan Goldin visited El Paso and our > NASA-sponsored > > > center, > > > > I was impressed by the fact that he did realize the necessity to use > > > > intelligent methods. I know that this realization is the result of a > hard > > > > education work, persistence, persistence, and persistence. I did give > my > > > spiel > > > > about intervals then, but of course, it is not enough. > > > > > > > > We need to use every effort to educate NASA folks and maybe, they will > > > start > > > > understanding the need for interval computations as well. The best way > to > > > > educate them is to show that it works on their real-life problems. I > > think > > > > control people have a good chance of using interval-based robust > control, > > > etc. > > > > With the new worry, we have a slightly better chance to get our foot > in. > > > > > > > > Vladik > > > > > > > > > From: "Eldar Musayev \(usa.net\)" > > > > > To: "Vladik Kreinovich" , > > > > > > > > > Subject: Re: opportunity ? > > > > > Date: Sun, 29 Apr 2001 12:16:40 -0700 > > > > > MIME-Version: 1.0 > > > > > Content-Transfer-Encoding: 7bit > > > > > X-Priority: 3 > > > > > X-MSMail-Priority: Normal > > > > > X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4133.2400 > > > > > > > > > > Generally, no. > > > > > > > > > > NASA problems are usually not with the floating-point arithmetic, > but > > > with > > > > > generic software bugs, which call for a new level of theoritization > in > > > > > Computer Science about how to write reliable (no generic bugs, not a > > > > > computation per se) programs, how to "control those stupid > programmers > > > > > without doing the job for them" (which is, luckily for researchers, > > > anyway > > > > > impossible both ways), with grants for researchers in these areas, > with > > > new > > > > > Rational (and probably Irrational too) Roses etc. And yes, it's a > lot > > of > > > > > sarcasm about such techniques, though not all of them. I think that > as > > a > > > > > direct result, you may expect a burst of publications on XP > > programming, > > > so > > > > > if you can link your research with XP programming, it may be a > > > reasonable > > > > > shot. There could be some other prospective areas, but they seems to > be > > > even > > > > > farther from interval arithmetic. So otherwise, chances seems to be > > dim. > > > > > > > > > > -- Eldar > > > > > > > > > > > > > > > ----- Original Message ----- > > > > > From: "Vladik Kreinovich" > > > > > To: > > > > > Sent: Friday, April 27, 2001 4:57 PM > > > > > Subject: opportunity ? > > > > > > > > > > > > > > > > > > > > > > "NASA Explores Future of Software" > > > > > > Wall Street Journal (04/26/01) P. A1; Wessel, David > > > > > > > > > > > > NASA administrator Daniel Goldin is planning to collaborate > > with > > > > > > universities and businesses to develop more reliable > > > > > > software at the Ames Research Center in Silicon Valley. > Along > > > with > > > > > the > > > > > > military, NASA has historically led in software > > > > > > development for its space projects. As business software > > > matured, > > > > > the > > > > > > focus of NASA and the corporate sector > > > > > > diverged more clearly, but Goldin believes that they may be > > > drawing > > > > > near > > > > > > again. Henry McDonald, the NASA physicist > > > > > > in charge of the Ames center, says that applications need to > be > > > > > found > > > > > > that support both business and NASA > > > > > > functions. He says that both large financial firms and NASA > > need > > > > > more > > > > > > reliable software to manage reams of digital > > > > > > information flowing continuously. Goldin is aware that NASA > has > > > > > suffered > > > > > > software-wise in recent years, and that > > > > > > better programs could help prevent disasters like the crash > of > > a > > > > > Mars > > > > > > Polar Lander in late 1999. NASA has already > > > > > > succeeded in signing on one research partner in IBM, who > shares > > > the > > > > > view > > > > > > that such collaborative research can > > > > > > improve both the process and the end product. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > ____________________________________________________________________ > > Get free email and a permanent address at http://www.netaddress.com/?N=1 > > > ____________________________________________________________________ Get free email and a permanent address at http://www.netaddress.com/?N=1 From owner-reliable_computing [at] interval [dot] usl.edu Sun Apr 29 21:03:46 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id VAA10355 for reliable_computing-outgoing; Sun, 29 Apr 2001 21:03:45 -0500 (CDT) Received: from interval.usl.edu (rbk5287@interval [130.70.43.77]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with SMTP id VAA10350 for ; Sun, 29 Apr 2001 21:03:41 -0500 (CDT) Message-Id: <200104300203.VAA10350 [at] interval [dot] usl.edu> Date: Sun, 29 Apr 2001 21:03:41 -0500 (CDT) From: "Kearfott R. Baker" Reply-To: "Kearfott R. Baker" Subject: Re: Reliable computing in the high performance environment To: reliable_computing [at] interval [dot] louisiana.edu MIME-Version: 1.0 Content-Type: TEXT/plain; charset=us-ascii Content-MD5: AZZskhcQ4Z4rSptj6AkR9A== X-Mailer: dtmail 1.2.1 CDE Version 1.2.1 SunOS 5.6 sun4m sparc Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk ------------- Begin Forwarded Message ------------- Date: Fri, 27 Apr 2001 22:51:25 -0500 (EST) From: ChaoYang Gau To: owner-reliable_computing [at] interval [dot] usl.edu Subject: Re: Reliable computing in the high performance environment > Dear interval researchers, > > I am working with Prof. Dalcidio Claudio and we are looking for works and > libries which works with reliable > computing in the high performance environment (cluster environment, use > interval in MPI or PVM)). > I'm looking for references about this topics, researches groups and interval > libraries (projects > similar to INTLAB, PROFIL, XSC-languagens, FORTE Fortran HPC, VLSI > Coprocessor for Accurate Vector Arithmetic, etc...). > > Best regards, > > Carlos Holbig Dear Carlos Hoblig, Some references regarding the parallel implementation of interval analysis on a cluster of workstations are available at http://www.nd.edu/~cgau/hpc.html C.-Y. Gau, and M. A. Stadtherr, "Parallel Branch-and-Bound for Chemical Engineering Applications: Load Balancing and Scheduling Issues," to appear in Lecture Notes in Computer Science (2000). C.-Y. Gau, M. A. Stadtherr, "A Systematic Analysis of Dynamic Load Balancing Strategies for Parallel Interval Analysis," AIChE Annual Meeting, Dallas, TX, October 31 - November 5, 1999. SLIDES Regards, Tony ------------------------------------------------------------------------------ Tony (Chao-Yang) Gau cgau [at] nd [dot] edu. Department of Chemical Engineering, University of Notre Dame Notre Dame, IN 46556 http://www.nd.edu/~cgau (URL) ------------------------------------------------------------------------------ ------------- End Forwarded Message ------------- --------------------------------------------------------------- R. Baker Kearfott, rbk [at] louisiana [dot] edu (337) 482-5346 (fax) (337) 482-5270 (work) (337) 981-9744 (home) URL: http://interval.louisiana.edu/kearfott.html Department of Mathematics, University of Louisiana at Lafayette Box 4-1010, Lafayette, LA 70504-1010, USA --------------------------------------------------------------- From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 30 05:00:02 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id FAA11366 for reliable_computing-outgoing; Mon, 30 Apr 2001 05:00:01 -0500 (CDT) Received: from relay.kfupm.edu.sa (relay.kfupm.edu.sa [212.26.1.21]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id EAA11350 for ; Mon, 30 Apr 2001 04:59:45 -0500 (CDT) Received: from relayin.kfupm.edu.sa (relayin.kfupm.edu.sa [212.26.2.26]) by relay.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id MAA69906; Mon, 30 Apr 2001 12:48:02 +0300 Received: from iws20.dpc.kfupm.edu.sa (iws20.dpc.kfupm.edu.sa [196.15.32.20]) by relayin.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id MAA14876; Mon, 30 Apr 2001 12:46:17 +0300 Received: from itcH70.itc.kfupm.edu.sa (itc.itc.kfupm.edu.sa [196.15.32.8]) by iws20.dpc.kfupm.edu.sa (8.9.2/8.9.2) with ESMTP id MAA41580; Mon, 30 Apr 2001 12:39:40 +0400 Received: from smtp (webnews.kfupm.edu.sa [196.15.32.251] (may be forged)) by itcH70.itc.kfupm.edu.sa (AIX4.3/8.9.3/8.9.2) with SMTP id MAA90814; Mon, 30 Apr 2001 12:55:23 +0300 Received: from 196.1.65.140 by smtp (InterScan E-Mail VirusWall NT); Mon, 30 Apr 2001 12:57:17 +0300 Message-ID: <000b01c0d15b$e95d5150$8c4101c4@icssarfraz> From: "Dr. Muhammad Sarfraz" To: "Dr. Muhammad Sarfraz" Subject: GDAM_CGIM'2001 Date: Mon, 30 Apr 2001 12:57:00 +0300 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0006_01C0D175.0C1C8B70" X-Priority: 1 X-MSMail-Priority: High X-Mailer: Microsoft Outlook Express 5.00.2919.6700 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk This is a multi-part message in MIME format. ------=_NextPart_000_0006_01C0D175.0C1C8B70 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable APOLOGIES TO THOSE WHO RECEIVE THIS CALL MORE THAN ONCE PLEASE NOTICE THAT THE DEADLINE FOR SUBMISSION OF PAPERS IS EXTENDED TO May 07, 2001 Special Session on=20 GEOMETRIC DESIGN & Modeling The International Conference on COMPUTER GRAPHICS AND IMAGING=20 (CGIM 2001) AUGUST 13-16, 2001 Honolulu, Hawaii, USA. Aims and Scope Computer Aided (CA) techniques for Geometric Design (GD) are extremely = useful tools in a number of industrial settings. Specifically, the = application of geometry to current problems of designing and = manufacturing have grown at a large scale. In addition to its critical = importance in the traditional fields of automobile and aircraft = manufacturing, shipbuilding, and general product design, more recently, = the CAGD methods have also proven to be indispensable in a variety of = modern industries, including computer vision, robotics, medical imaging, = visualization, and even media. The purpose of the Geometric Design session in this conference = (CGIM'2001) is to focus on interdisciplinary methods and affiliate = research done among various disciplines with the goal of stimulating = views and providing a forum where researchers and practitioners can = discuss the latest developments linked to Computer Graphics and Imaging. = Papers presenting original research with the theme of "Geometric Design = & Modeling" are being sought. Suggested topics include (but are not = limited to):=20 *=20 * Curve and Surface Design * Visualization * Computer Graphics * Geometric Modeling=20 * Geometric Algorithms=20 * Manufacturing Geometry * Computational Geometry * Reverse Engineering * Solid Modeling * Computer Aided Design (CAD) * Computer Aided Manufacturing (CAM)=20 * Font Design * Multiresolution Methods * Geometric Transformations * Mesh Processing * Shape Abstraction and Modeling=20 * Shape Analysis and Description * Industrial Applications=20 IMPORTANT DEADLINES Submissions due May 07, 2001 Notification of acceptance May 30, 2001 Registration, full payment, and final papers due June 30, 2001 All Submissions should be addressed to: Chair: GDAM-CGIM'2001 Dr. M. Sarfraz Department of Information and Computer Science, King Fahd University of Petroleum and Minerals, KFUPM # 1510, Dhahran 31261. Saudi Arabia. Email: sarfraz [at] kfupm [dot] edu.sa For more information, please visit: = http://www.iasted.com/conferences/2001/hawaii/cgim.htm ------=_NextPart_000_0006_01C0D175.0C1C8B70 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable *******************************

APOLOGIES TO THOSE WHO RECEIVE THIS CALL MORE THAN = ONCE

PLEASE NOTICE THAT THE DEADLINE FOR = SUBMISSION OF PAPERS IS EXTENDED TO

May 07, 2001

Special = Session on

GEOMETRIC DESIGN & Modeling

The International Conference on COMPUTER GRAPHICS AND IMAGING =

(CGIM 2001)

AUGUST 13-16, 2001

Honolulu, Hawaii, USA.

 

Aims and = Scope

Computer Aided (CA) techniques for Geometric Design (GD) are extremely useful tools in a number of industrial settings. Specifically, = the application of geometry to current problems of designing and = manufacturing have grown at a large scale. In addition to its critical importance in the traditional fields of automobile and aircraft manufacturing, = shipbuilding, and general product design, more recently, the CAGD methods have also proven = to be indispensable in a variety of modern industries, including computer vision, robotics, = medical imaging, visualization, and even media.

 

The purpose of the Geometric Design session in this = conference (CGIM’2001) is to focus on interdisciplinary methods and affiliate = research done among various disciplines with the goal of stimulating views and = providing a forum where researchers and practitioners can discuss the latest = developments linked to Computer Graphics and Imaging.

 

Papers presenting original research with the theme of "Geometric Design & Modeling" are being sought. Suggested = topics include (but are not limited to):

 


=DC    = Curve and Surface Design

=DC    = Visualization=

=DC    = Computer Graphics

=DC    = Geometric Modeling

=DC    = Geometric Algorithms

=DC    = Manufacturing Geometry

=DC    = Computational Geometry

=DC    = Reverse Engineering

=DC    = Solid Modeling

=DC    = Computer Aided Design (CAD)

=DC    = Computer Aided Manufacturing (CAM)

=DC    = Font Design

=DC    = Multiresolution Methods

=DC    = Geometric Transformations

=DC    = Mesh Processing

=DC    = Shape Abstraction and Modeling

=DC    = Shape Analysis and Description

=DC    = Industrial Applications


 

IMPORTANT DEADLINES

Submissions due        = ;           =         = ;            =             = ;            = ;    May 07, 2001

Notification of acceptance               = ;            =             = ;            = ;    May 30, 2001

Registration, full payment, and final = papers due         = ;            = ;    June 30, = 2001

 

All Submissions should = be addressed to:

Chair: = GDAM-CGIM’2001

Dr. M. = Sarfraz

Department of Information and Computer Science,

King Fahd University of Petroleum and Minerals,

KFUPM # 1510, Dhahran 31261.

Saudi Arabia.

Email:    sarfraz [at] kfupm [dot] edu.sa

For more information, please visit: http://ww= w.iasted.com/conferences/2001/hawaii/cgim.htm

------=_NextPart_000_0006_01C0D175.0C1C8B70-- From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 30 05:56:05 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id FAA11730 for reliable_computing-outgoing; Mon, 30 Apr 2001 05:56:05 -0500 (CDT) Received: from postmarq.mu.edu (maia.mu.edu [134.48.1.6]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id FAA11725 for ; Mon, 30 Apr 2001 05:56:01 -0500 (CDT) Received: from Marquette.edu ([134.48.24.8]) by postmarq.mu.edu (Netscape Messaging Server 4.15) with ESMTP id GCLQD800.8EJ; Mon, 30 Apr 2001 05:55:56 -0500 Message-ID: <3AED44B8.F3202E78 [at] Marquette [dot] edu> Date: Mon, 30 Apr 2001 05:55:52 -0500 From: George Corliss Reply-To: George.Corliss [at] Marquette [dot] edu Organization: Marquette University X-Mailer: Mozilla 4.7 [en] (Win98; U) X-Accept-Language: en MIME-Version: 1.0 To: reliable_computing [at] interval [dot] louisiana.edu CC: "Dr. George Corliss" Subject: Re: opportunity ? Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Perhaps marginally relevant: Another possible contribution of interval techniques is to debugging. 1. Some 15 years ago, Louis Rall and I developed a package for numerical quadrature using interval arithmetic. Among our test problems were several integrals of the general form Integral_a^b asin (sin (x)) - x dx whose answers (on suitable intervals) are ZERO. I think we uncovered some 30 errors in our code (ok, I should be a more careful programmer) where the program returned "validated enclosures" of the form [-2, -1] * 10^{-17}. In floating point arithmetic, -2*10^{-17} and -1*10^{-17} are pretty good approximations to double precision zero. However, as an interval, they do not enclose zero, disclosing a coding error. Without intervals, those errors would have gone undetected. 2. At about the same time, I worked with a scientist at Amoco doing oil reservoir simulation, specifically studying the propagation of a crack in the rock as water is pumped into a well. His program crashed with SQRT( negative ) after some 118 seemingly well behaved time steps. He could find no error. We used an interval version of a subalgorithm to discover that the way he was formulating an initial guess for a Newton iteration was not sufficiently accurate as the crack became sufficiently long. Once he identified the cause of the SQRT( negative ) crash, he could easily fix it by using a more accurate Newton iteration initialization. The eventual program had no interval methods. We only used the interval techniques for debugging. Without intervals, this reasonably bright guy had been unable to locate his error in several months of study. Take home: Interval techniques can sometimes render otherwise hidden error obvious. Dr. George F. Corliss Dept. Math, Stat, Comp Sci Marquette University P.O. Box 1881 Milwaukee, WI 53201-1881 USA George.Corliss [at] Marquette [dot] edu Office: 414-288-6599; Dept: 288-7573; Fax: 288-5472 From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 30 09:11:53 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id JAA00386 for reliable_computing-outgoing; Mon, 30 Apr 2001 09:11:53 -0500 (CDT) Received: from mail.comset.net (IDENT:root [at] mail [dot] comset.net [213.172.7.103]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id JAA00381 for ; Mon, 30 Apr 2001 09:11:47 -0500 (CDT) Received: from vyachesn (3-168.dialup.comset.net [213.172.3.168]) by mail.comset.net (8.9.3/2.0.1/ComSet) with ESMTP id SAA21979 for ; Mon, 30 Apr 2001 18:11:42 +0400 Message-ID: <001e01c0d17f$570570c0$a803acd5@vyachesn> From: "Vyacheslav Nesterov" To: Subject: Reliable Computing, issue 5, volume 7 Date: Mon, 30 Apr 2001 10:32:10 +0400 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_0064_01C0D160.D0601FC0" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.2615.200 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2615.200 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk This is a multi-part message in MIME format. ------=_NextPart_000_0064_01C0D160.D0601FC0 Content-Type: text/plain; charset="koi8-r" Content-Transfer-Encoding: quoted-printable Reliable Computing Volume 7, issue 5, 2001 Mathematical Research Idempotent Interval Analysis and Optimization Problems Grigori L. Litvinov, Andrei N. Sobolevskii 353-377 Verified High-Order Inversion of Functional Dependencies=20 and Interval Newton Methods Martin Berz, Jens Hoefkens 379-398 Cheap and Tight Bounds on the Solution Set of Perturbed Systems of Nonlinear Equations Lubomir V. Kolev, Ivo P. Nenov 399-408 On Maximal Inner Estimation of the Solution Sets of Linear Systems with Interval Parameters Irene A. Sharaya 409-424 Short Communication An Example of Singularity in Nonlinear Systems R. Baker Kearfott 425-429 -- Claude E. Shannon (1916--2001) 431-432 ------=_NextPart_000_0064_01C0D160.D0601FC0 Content-Type: text/html; charset="koi8-r" Content-Transfer-Encoding: quoted-printable
          &nbs= p;    =20 Reliable=20 Computing
          =    =20 Volume 7, issue 5, 2001
 
          &nbs= p;    =20 Mathematical Research
 
Idempotent Interval Analysis and = Optimization=20 Problems
Grigori L. Litvinov, Andrei N. = Sobolevskii
353-377
 
Verified High-Order Inversion of = Functional=20 Dependencies
and Interval Newton Methods
Martin Berz, Jens=20 Hoefkens
379-398
 
Cheap and Tight Bounds on the = Solution Set of=20 Perturbed Systems of
Nonlinear Equations
Lubomir V. Kolev, Ivo P.=20 Nenov
399-408
 
On Maximal Inner Estimation of = the Solution=20 Sets of Linear Systems with
Interval Parameters
Irene A.=20 Sharaya
409-424
 
          &nbs= p;       =20 Short Communication
 
An Example of Singularity in = Nonlinear=20 Systems
R. Baker Kearfott
425-429
 
--
Claude E. Shannon=20 (1916--2001)
431-432
------=_NextPart_000_0064_01C0D160.D0601FC0-- From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 30 09:25:52 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id JAA00673 for reliable_computing-outgoing; Mon, 30 Apr 2001 09:25:52 -0500 (CDT) Received: from Post-Office.UH.EDU (Post-Office.UH.EDU [129.7.1.20]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id JAA00668 for ; Mon, 30 Apr 2001 09:25:49 -0500 (CDT) Received: from CONVERSION-DAEMON by Post-Office.UH.EDU (PMDF V5.2-32 #40812) id <0GCM0080102HHT@Post-Office.UH.EDU> for reliable_computing [at] interval [dot] louisiana.edu; Mon, 30 Apr 2001 09:25:37 -0500 (CDT) Received: from contact.uh.edu (Contact.UH.EDU [129.7.69.1]) by Post-Office.UH.EDU (PMDF V5.2-32 #40812) with ESMTP id <0GCM0076D02HSX@Post-Office.UH.EDU>; Mon, 30 Apr 2001 09:25:30 -0500 (CDT) Received: by Contact.UH.EDU with Internet Mail Service (5.5.2650.21) id <25XYJDCR>; Mon, 30 Apr 2001 09:25:29 -0500 Content-return: allowed Date: Mon, 30 Apr 2001 09:23:37 -0500 From: "Hu, Chenyi" Subject: RE: Reliable computing in the high performance environment To: =?iso-8859-1?Q?=27Carlos_Amaral_H=F6lbig=27?= , reliable_computing [at] interval [dot] louisiana.edu Message-id: MIME-version: 1.0 X-Mailer: Internet Mail Service (5.5.2650.21) Content-type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by interval.usl.edu id JAA00669 Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk Yes, we have made parallel implementations on the nonlinear global optimization package GlobSol with MPI.Please refer the URL http://happy.dt.uh.edu/~sun/ParaGlobSol.html Chenyi Hu; Ph.D Computer and Mathematical Sciences University of Houston-Downtown One Main Street Houston, Texas, 77002 Phone: 713 221 8414 Fax: 713 221 8086 E-mail: CHu [at] uh [dot] edu http://happy.dt.uh.edu/~hu/Hu.html -----Original Message----- From: Carlos Amaral Hölbig [mailto:holbig [at] upf [dot] tche.br] Sent: Thursday, April 26, 2001 11:32 AM To: reliable_computing [at] interval [dot] louisiana.edu Subject: Reliable computing in the high performance environment Dear interval researchers, I am working with Prof. Dalcidio Claudio and we are looking for works and libries which works with reliable computing in the high performance environment (cluster environment, use interval in MPI or PVM)). I'm looking for references about this topics, researches groups and interval libraries (projects similar to INTLAB, PROFIL, XSC-languagens, FORTE Fortran HPC, VLSI Coprocessor for Accurate Vector Arithmetic, etc...). Best regards, Carlos Holbig _____________________________________________ Prof. Msc. Carlos Amaral Hölbig Passo Fundo University - Brazil E-mail: holbig [at] upf [dot] tche.br http://vitoria.upf.tche.br/~holbig From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 30 13:33:01 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id NAA01192 for reliable_computing-outgoing; Mon, 30 Apr 2001 13:33:00 -0500 (CDT) Received: from zcars0m9.nortelnetworks.com (h157s242a129n47.user.nortelnetworks.com [47.129.242.157]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with SMTP id NAA01186 for ; Mon, 30 Apr 2001 13:32:53 -0500 (CDT) Received: from zcars04e.nortelnetworks.com by zcars0m9.nortelnetworks.com (SMI-8.6/SMI-SVR4) id OAA05117; Mon, 30 Apr 2001 14:32:08 -0400 Received: from zcard015.ca.nortel.com (actually zcard015) by zcars04e.nortelnetworks.com; Mon, 30 Apr 2001 14:31:59 -0400 Received: by zcard015.ca.nortel.com with Internet Mail Service (5.5.2653.19) id ; Mon, 30 Apr 2001 14:31:59 -0400 Message-ID: <03E3E0690542D211A1490000F80836F405B51376 [at] zcard00f [dot] ca.nortel.com> From: "Bill Older" To: George.Corliss [at] Marquette [dot] edu, reliable_computing [at] interval [dot] louisiana.edu Subject: RE: opportunity ? Date: Mon, 30 Apr 2001 14:31:54 -0400 MIME-Version: 1.0 X-Mailer: Internet Mail Service (5.5.2653.19) Content-Type: multipart/alternative; boundary="----_=_NextPart_001_01C0D1A3.D4BCC7B0" X-Orig: Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk This message is in MIME format. Since your mail reader does not understand this format, some or all of this message may not be legible. ------_=_NextPart_001_01C0D1A3.D4BCC7B0 Content-Type: text/plain; charset="iso-8859-1" Debugging Stories: I am not sure whether this would be considered relevant to the original discussion or not, but perhaps it ought to be. In the early 90's I was working on an interval based "logical arithmetic" ( interval constraints) implemented in BNR Prolog. The implementation at that time was on 68K Macs. The "relational interval arithmetic" operations were written in 68K assembler to access the Motorola math coprocessor interface based on IEEE 784 ( actually quite a nice interface to work with for this purpose). On top of all the real constraint stuff I had written some application programs to do non-linear constrained ( >=) optimizations based on Lagrangians. I had a set of modest sized non-linear optimizations which I used as test cases. These had been tested previously on a Mac in which the FPU was emulated in s/w. When I got a Mac upgrade to a new & faster machine with a real FPU I was anxious to run some benchmarks (understandably). Much to my surprise one of the test cases now gave a different answer-- "NO". As this was a well-posed optimization problem this was not really an acceptable answer. I originally suspected some problem in the high level software, either the prolog application or the constraint set up code. But dumping the actual constraint network showed the same structure as in the previous machine. I then modified the test program to non-deterministically subset the constraints until it found a minimal failing subproblem. I then tried several hand-crafted versions of this minimal problem written in various ways; eventually I got down to a single expression which failed when it should not have. When I single-stepped through this expression ( one contraint operation at a time), it worked however, but would fail at full speed. So I began to suspect a h/w problem. I trapped the code at the first failing operation, and examined the values of the intervals in memory and in the fpu registers and discovered a slight difference in one variable - the least significant bit of the mantissa ( in the fpu's internal 96 bit fp representation)was not the same. After much fiddling about it appeared that depending on the contents of the fpu registers, the last few bits of the float value might not get latched properly during transfer to the fpu register because of relative timing issues. This information along with the relevant test cases etc was passed along to Motorola ( and getting through to some one who understood was a VERY long process) and a couple of months later I got some new fpu chips that worked properly. (At least for this suite of problems!) I figure this story has just got to have some sort of a moral, but have never been able to figure out just what it is! -----Original Message----- From: George Corliss [mailto:George.Corliss [at] Marquette [dot] edu] Sent: Monday, April 30, 2001 6:56 AM To: reliable_computing [at] interval [dot] louisiana.edu Cc: Dr. George Corliss Subject: Re: opportunity ? Perhaps marginally relevant: Another possible contribution of interval techniques is to debugging. 1. Some 15 years ago, Louis Rall and I developed a package for numerical quadrature using interval arithmetic. Among our test problems were several integrals of the general form Integral_a^b asin (sin (x)) - x dx whose answers (on suitable intervals) are ZERO. I think we uncovered some 30 errors in our code (ok, I should be a more careful programmer) where the program returned "validated enclosures" of the form [-2, -1] * 10^{-17}. In floating point arithmetic, -2*10^{-17} and -1*10^{-17} are pretty good approximations to double precision zero. However, as an interval, they do not enclose zero, disclosing a coding error. Without intervals, those errors would have gone undetected. 2. At about the same time, I worked with a scientist at Amoco doing oil reservoir simulation, specifically studying the propagation of a crack in the rock as water is pumped into a well. His program crashed with SQRT( negative ) after some 118 seemingly well behaved time steps. He could find no error. We used an interval version of a subalgorithm to discover that the way he was formulating an initial guess for a Newton iteration was not sufficiently accurate as the crack became sufficiently long. Once he identified the cause of the SQRT( negative ) crash, he could easily fix it by using a more accurate Newton iteration initialization. The eventual program had no interval methods. We only used the interval techniques for debugging. Without intervals, this reasonably bright guy had been unable to locate his error in several months of study. Take home: Interval techniques can sometimes render otherwise hidden error obvious. Dr. George F. Corliss Dept. Math, Stat, Comp Sci Marquette University P.O. Box 1881 Milwaukee, WI 53201-1881 USA George.Corliss [at] Marquette [dot] edu Office: 414-288-6599; Dept: 288-7573; Fax: 288-5472 ------_=_NextPart_001_01C0D1A3.D4BCC7B0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable RE: opportunity ?

Debugging Stories:

I am not sure whether this would be considered = relevant to the original discussion
or not, but perhaps it ought to be.

In the early 90's I was working on an interval based = "logical arithmetic"
( interval constraints) implemented in BNR Prolog. = The implementation at that time
was on 68K Macs.  The "relational interval = arithmetic" operations were written in
68K assembler to access the Motorola math = coprocessor interface based on IEEE 784
( actually quite a nice interface to work with for = this purpose). 

On top of all the real constraint stuff I had written = some application programs
to do non-linear constrained ( >=3D) = optimizations based on Lagrangians.  I had a set of
modest sized non-linear optimizations which I used = as test cases. These had been
tested previously on a Mac in which the FPU was = emulated in s/w.

When I got a Mac upgrade to a new & faster = machine with a real FPU I was anxious
to run some benchmarks (understandably).  Much = to my surprise one of the test cases
now gave a different answer-- = "NO".   As this was a well-posed optimization = problem this was
not really an acceptable answer.

I originally suspected some problem in the high level = software, either the prolog
application or the constraint set up code. But = dumping the actual constraint network showed the same structure as in = the previous machine.  I then modified the test program = to

non-deterministically subset the constraints until it = found a minimal failing subproblem.
I then tried several hand-crafted versions of this = minimal problem written in various ways;
eventually I got down to a single expression which = failed when it should not have.

When I single-stepped through this expression ( one = contraint operation at a time),
it worked however, but would fail at full = speed.  So I began to suspect a h/w problem.
I trapped the code at the first failing operation, = and examined the values of the
intervals in memory and in the fpu registers and = discovered a slight difference in one variable - the least significant = bit of the mantissa ( in the fpu's internal 96 bit fp = representation)was not the same. 

After much fiddling about it appeared that depending = on the contents of the fpu registers,
the last few bits of the float value might not get = latched properly during transfer to the
fpu register because of relative timing = issues. 

This information along with the relevant test cases = etc was passed along to Motorola ( and getting through to some one who = understood was a VERY long process) and a couple of months

later I got some new fpu chips that worked properly. = (At least for this suite of problems!)

I figure this story has just got to have some sort of = a moral, but have never been able to figure out just what it is!




 



-----Original Message-----
From: George Corliss [mailto:George.Corliss@Marqu= ette.edu]
Sent: Monday, April 30, 2001 6:56 AM
To: reliable_computing [at] interval [dot] louisiana.edu
Cc: Dr. George Corliss
Subject: Re: opportunity ?


Perhaps marginally relevant:

Another possible contribution of interval techniques = is
to debugging.

1.  Some 15 years ago, Louis Rall and I = developed a package
for numerical quadrature using interval = arithmetic.  Among
our test problems were several integrals of the = general form
   Integral_a^b  asin (sin (x)) - = x  dx
whose answers (on suitable intervals) are = ZERO.  I think
we uncovered some 30 errors in our code (ok, I = should be
a more careful programmer) where the program = returned
"validated enclosures" of the form [-2, = -1] * 10^{-17}.
In floating point arithmetic, -2*10^{-17} and = -1*10^{-17}
are pretty good approximations to double precision = zero.
However, as an interval, they do not enclose zero, = disclosing
a coding error. 

Without intervals, those errors would have gone = undetected.

2.  At about the same time, I worked with a = scientist at Amoco
doing oil reservoir simulation, specifically = studying the
propagation of a crack in the rock as water is = pumped into
a well.  His program crashed with SQRT( = negative ) after some
118 seemingly well behaved time steps.  He = could find no
error.  We used an interval version of a = subalgorithm to
discover that the way he was formulating an initial = guess
for a Newton iteration was not sufficiently accurate = as the
crack became sufficiently long.  Once he = identified the
cause of the SQRT( negative ) crash, he could = easily
fix it by using a more accurate Newton iteration =
initialization.  The eventual program had no = interval
methods.  We only used the interval techniques = for debugging.

Without intervals, this reasonably bright guy had = been
unable to locate his error in several months of = study.


Take home: Interval techniques can sometimes render = otherwise
hidden error obvious.

Dr. George F. Corliss
Dept. Math, Stat, Comp Sci
Marquette University
P.O. Box 1881
Milwaukee, WI  53201-1881  USA
George.Corliss [at] Marquette [dot] edu
Office: 414-288-6599;  Dept: 288-7573; Fax: = 288-5472

------_=_NextPart_001_01C0D1A3.D4BCC7B0-- From owner-reliable_computing [at] interval [dot] usl.edu Mon Apr 30 14:09:50 2001 Received: (from root@localhost) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) id OAA01519 for reliable_computing-outgoing; Mon, 30 Apr 2001 14:09:50 -0500 (CDT) Received: from marnier.ucs.louisiana.edu (root@[130.70.40.2]) by interval.usl.edu (8.9.1/8.9.1/interval-math-majordomo-1.1) with ESMTP id OAA01514 for ; Mon, 30 Apr 2001 14:09:47 -0500 (CDT) Received: from liberty (liberty.louisiana.edu [130.70.46.171]) by marnier.ucs.louisiana.edu (8.9.1/8.9.1/ucs-mx-host_1.3-virus) with SMTP id NAA27503; Mon, 30 Apr 2001 13:47:37 -0500 (CDT) Message-Id: <2.2.32.20010430184935.012ff058 [at] 130 [dot] 70.132.231> X-Sender: rbk5287 [at] 130 [dot] 70.132.231 X-Mailer: Windows Eudora Pro Version 2.2 (32) Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Date: Mon, 30 Apr 2001 13:49:35 -0500 To: "Bill Older" , reliable_computing [at] interval [dot] louisiana.edu From: "R. Baker Kearfott" Subject: RE: opportunity ? Sender: owner-reliable_computing [at] interval [dot] usl.edu Precedence: bulk The way I read your description of your debugging process, I understand that the hardware error would not have been discovered if you had not been working with intervals. By the way, in the mid 1990's, two of our Ph.D. students trained in interval analysis got positions at Motorola :-) Best regards, Baker At 02:31 PM 4/30/01 -0400, Bill Older wrote: > >Debugging Stories: > >I am not sure whether this would be considered relevant to the original >discussion >or not, but perhaps it ought to be. > >In the early 90's I was working on an interval based "logical arithmetic" >( interval constraints) implemented in BNR Prolog. The implementation at >that time >was on 68K Macs. The "relational interval arithmetic" operations were >written in >68K assembler to access the Motorola math coprocessor interface based on >IEEE 784 >( actually quite a nice interface to work with for this purpose). > >On top of all the real constraint stuff I had written some application >programs >to do non-linear constrained ( >=) optimizations based on Lagrangians. I >had a set of >modest sized non-linear optimizations which I used as test cases. These had >been >tested previously on a Mac in which the FPU was emulated in s/w. > >When I got a Mac upgrade to a new & faster machine with a real FPU I was >anxious >to run some benchmarks (understandably). Much to my surprise one of the >test cases >now gave a different answer-- "NO". As this was a well-posed optimization >problem this was >not really an acceptable answer. > >I originally suspected some problem in the high level software, either the >prolog >application or the constraint set up code. But dumping the actual constraint >network showed the same structure as in the previous machine. I then >modified the test program to >non-deterministically subset the constraints until it found a minimal >failing subproblem. >I then tried several hand-crafted versions of this minimal problem written >in various ways; >eventually I got down to a single expression which failed when it should not >have. > >When I single-stepped through this expression ( one contraint operation at a >time), >it worked however, but would fail at full speed. So I began to suspect a >h/w problem. >I trapped the code at the first failing operation, and examined the values >of the >intervals in memory and in the fpu registers and discovered a slight >difference in one variable - the least significant bit of the mantissa ( in >the fpu's internal 96 bit fp representation)was not the same. > >After much fiddling about it appeared that depending on the contents of the >fpu registers, >the last few bits of the float value might not get latched properly during >transfer to the >fpu register because of relative timing issues. > >This information along with the relevant test cases etc was passed along to >Motorola ( and getting through to some one who understood was a VERY long >process) and a couple of months >later I got some new fpu chips that worked properly. (At least for this >suite of problems!) > >I figure this story has just got to have some sort of a moral, but have >never been able to figure out just what it is! > > > > > > > > >-----Original Message----- >From: George Corliss [mailto:George.Corliss [at] Marquette [dot] edu] >Sent: Monday, April 30, 2001 6:56 AM >To: reliable_computing [at] interval [dot] louisiana.edu >Cc: Dr. George Corliss >Subject: Re: opportunity ? > > >Perhaps marginally relevant: > >Another possible contribution of interval techniques is >to debugging. > >1. Some 15 years ago, Louis Rall and I developed a package >for numerical quadrature using interval arithmetic. Among >our test problems were several integrals of the general form > Integral_a^b asin (sin (x)) - x dx >whose answers (on suitable intervals) are ZERO. I think >we uncovered some 30 errors in our code (ok, I should be >a more careful programmer) where the program returned >"validated enclosures" of the form [-2, -1] * 10^{-17}. >In floating point arithmetic, -2*10^{-17} and -1*10^{-17} >are pretty good approximations to double precision zero. >However, as an interval, they do not enclose zero, disclosing >a coding error. > >Without intervals, those errors would have gone undetected. > >2. At about the same time, I worked with a scientist at Amoco >doing oil reservoir simulation, specifically studying the >propagation of a crack in the rock as water is pumped into >a well. His program crashed with SQRT( negative ) after some >118 seemingly well behaved time steps. He could find no >error. We used an interval version of a subalgorithm to >discover that the way he was formulating an initial guess >for a Newton iteration was not sufficiently accurate as the >crack became sufficiently long. Once he identified the >cause of the SQRT( negative ) crash, he could easily >fix it by using a more accurate Newton iteration >initialization. The eventual program had no interval >methods. We only used the interval techniques for debugging. > >Without intervals, this reasonably bright guy had been >unable to locate his error in several months of study. > > >Take home: Interval techniques can sometimes render otherwise >hidden error obvious. > >Dr. George F. Corliss >Dept. Math, Stat, Comp Sci >Marquette University >P.O. Box 1881 >Milwaukee, WI 53201-1881 USA >George.Corliss [at] Marquette [dot] edu >Office: 414-288-6599; Dept: 288-7573; Fax: 288-5472 > > > > > >RE: opportunity ? > > >
> >

Debugging Stories: >

> >

I am not sure whether this would be considered relevant to the original discussion >
or not, but perhaps it ought to be. >

> >

In the early 90's I was working on an interval based "logical arithmetic" >
( interval constraints) implemented in BNR Prolog. The implementation at that time >
was on 68K Macs.  The "relational interval arithmetic" operations were written in >
68K assembler to access the Motorola math coprocessor interface based on IEEE 784 >
( actually quite a nice interface to work with for this purpose).  >

> >

On top of all the real constraint stuff I had written some application programs >
to do non-linear constrained ( >=) optimizations based on Lagrangians.  I had a set of >
modest sized non-linear optimizations which I used as test cases. These had been >
tested previously on a Mac in which the FPU was emulated in s/w. >

> >

When I got a Mac upgrade to a new & faster machine with a real FPU I was anxious >
to run some benchmarks (understandably).  Much to my surprise one of the test cases >
now gave a different answer-- "NO".   As this was a well-posed optimization problem this was >
not really an acceptable answer. >

> >

I originally suspected some problem in the high level software, either the prolog >
application or the constraint set up code. But dumping the actual constraint network showed the same structure as in the previous machine.  I then modified the test program to

> >

non-deterministically subset the constraints until it found a minimal failing subproblem. >
I then tried several hand-crafted versions of this minimal problem written in various ways; >
eventually I got down to a single expression which failed when it should not have. >

> >

When I single-stepped through this expression ( one contraint operation at a time), >
it worked however, but would fail at full speed.  So I began to suspect a h/w problem. >
I trapped the code at the first failing operation, and examined the values of the >
intervals in memory and in the fpu registers and discovered a slight difference in one variable - the least significant bit of the mantissa ( in the fpu's internal 96 bit fp representation)was not the same. 

> >

After much fiddling about it appeared that depending on the contents of the fpu registers, >
the last few bits of the float value might not get latched properly during transfer to the >
fpu register because of relative timing issues.  >

> >

This information along with the relevant test cases etc was passed along to Motorola ( and getting through to some one who understood was a VERY long process) and a couple of months

> >

later I got some new fpu chips that worked properly. (At least for this suite of problems!) >

> >

I figure this story has just got to have some sort of a moral, but have never been able to figure out just what it is! >

>
>
>
> >

  >

>
>
> >

-----Original Message----- >
From: George Corliss [mailto:George.Corliss [at] Marquette [dot] edu] >
Sent: Monday, April 30, 2001 6:56 AM >
To: reliable_computing [at] interval [dot] louisiana.edu >
Cc: Dr. George Corliss >
Subject: Re: opportunity ? >

>
> >

Perhaps marginally relevant: >

> >

Another possible contribution of interval techniques is >
to debugging. >

> >

1.  Some 15 years ago, Louis Rall and I developed a package >
for numerical quadrature using interval arithmetic.  Among >
our test problems were several integrals of the general form >
   Integral_a^b  asin (sin (x)) - x  dx >
whose answers (on suitable intervals) are ZERO.  I think >
we uncovered some 30 errors in our code (ok, I should be >
a more careful programmer) where the program returned >
"validated enclosures" of the form [-2, -1] * 10^{-17}. >
In floating point arithmetic, -2*10^{-17} and -1*10^{-17} >
are pretty good approximations to double precision zero. >
However, as an interval, they do not enclose zero, disclosing >
a coding error.  >

> >

Without intervals, those errors would have gone undetected. >

> >

2.  At about the same time, I worked with a scientist at Amoco >
doing oil reservoir simulation, specifically studying the >
propagation of a crack in the rock as water is pumped into >
a well.  His program crashed with SQRT( negative ) after some >
118 seemingly well behaved time steps.  He could find no >
error.  We used an interval version of a subalgorithm to >
discover that the way he was formulating an initial guess >
for a Newton iteration was not sufficiently accurate as the >
crack became sufficiently long.  Once he identified the >
cause of the SQRT( negative ) crash, he could easily >
fix it by using a more accurate Newton iteration >
initialization.  The eventual program had no interval >
methods.  We only used the interval techniques for debugging. >

> >

Without intervals, this reasonably bright guy had been >
unable to locate his error in several months of study. >

>
> >

Take home: Interval techniques can sometimes render otherwise >
hidden error obvious. >

> >

Dr. George F. Corliss >
Dept. Math, Stat, Comp Sci >
Marquette University >
P.O. Box 1881 >
Milwaukee, WI  53201-1881  USA >
George.Corliss [at] Marquette [dot] edu >
Office: 414-288-6599;  Dept: 288-7573; Fax: 288-5472 >

> > > --------------------------------------------------------------- R. Baker Kearfott, rbk [at] louisiana [dot] edu (337) 482-5346 (fax) (337) 482-5270 (work) (337) 981-9744 (home) URL: http://interval.louisiana.edu/kearfott.html Department of Mathematics, University of Louisiana at Lafayette Box 4-1010, Lafayette, LA 70504-1010, USA ---------------------------------------------------------------