next up previous notation contents
Next: 3.4.3 Optimality Up: 3.4 Polynomial Interval Arithmetic Previous: 3.4.1 Interpolating Polynomials

3.4.2 tex2html_wrap_inline34297 Charts

 

We now prove that the rules given in section gif, for constructing a tex2html_wrap_inline34297 chart, are correct.

The forbidden region is clearly correct since tex2html_wrap_inline31337 is a function; we simply decree that tex2html_wrap_inline37393 as our use of the tex2html_wrap_inline34297 chart does not depend on how such points are treated. For (x,y) in the zero region, tex2html_wrap_inline37399 and tex2html_wrap_inline37401 ; so for any point in the zero region tex2html_wrap_inline37403 , which implies tex2html_wrap_inline37405 .

For the remaining regions, consider the polynomial

math24682

From the construction of p(x) it is clear that

math24690

for any tex2html_wrap_inline37409 . Consider the polynomial

math24694

which interpolates tex2html_wrap_inline34275 for any value of m. The k roots of the k degree polynomial p are tex2html_wrap_inline37421 . The polynomial p has no other roots since it is not identically zero. For large x, p(x) is positive:

math24700

Imagine p(x) as x decreases; the sign of p(x) will reverse each time x crosses a root of p(x). This sign changing corresponds with the checkboard labelling of tex2html_wrap_inline34297 .

Consider the point (x,y) which is tex2html_wrap_inline37443 away from tex2html_wrap_inline34329 :

math24707

If

math24712

then

eqnarray24717

Earlier we proved q(x) interpolated tex2html_wrap_inline34275 for any m. We have now shown q(x) interpolates tex2html_wrap_inline34273 . The leading coefficient of q(x) is m. The sign of m relates to tex2html_wrap_inline37463 : the sign of m is positive if the region (x,y) resides in is labelled with tex2html_wrap_inline34287 ; the sign of m is negative if the region (x,y) resides in is labelled with tex2html_wrap_inline34289 .


next up previous notation contents
Next: 3.4.3 Optimality Up: 3.4 Polynomial Interval Arithmetic Previous: 3.4.1 Interpolating Polynomials
Jeff TupperMarch 1996