next up previous
Next: The Fritz-John Equations Up: On Interval Newton Methods Previous: On Interval Newton Methods

An Example of Interval Newton iteration

  Take
eqnarray260
with
displaymath832
An interval extension of the Jacobi matrix for f is
displaymath836
and its value at tex2html_wrap_inline64 is
displaymath840
The usual procedure (although not required in this special case) is to precondition the system
displaymath842
say, by the inverse of the midpoint matrix
displaymath1006
to obtain
displaymath846
i.e.
displaymath1010
The interval Gauss-Seidel method can then be used to compute sharper bounds on tex2html_wrap_inline850, beginning with tex2html_wrap_inline852. That is,
eqnarray325
Thus, the first component of tex2html_wrap_inline794 is
displaymath856
In the second step of the interval Gauss-Seidel method,
eqnarray331
so, rounded out to four digits, tex2html_wrap_inline794 is computed to be
displaymath860
This last inclusion proves that there exists a unique solution to F(x)=0 within tex2html_wrap_inline64, and hence, within tex2html_wrap_inline794. Furthermore, iteration of the procedure will result in bounds on the exact solution that become narrow quadratically.



R. Baker Kearfott
Wed May 27 11:22:02 CDT 1998