On 3/27/2012 5:05 PM, Edwin Woollett wrote:..
...
>
> To detect an odd function for a complicated function
> symbolically might not always be possible,
true
> but
> perhaps numerical sampling would be enough?
you could (probably) say that an integral's value is "bounded away from
zero"
or "can't say if it is bounded away from zero with the tools at our
disposal."
Another approach is to say that the quadrature program provides
only "absolute error" bounds if the answer is near zero, instead of
"relative error" (or X digits of accuracy) elsewhere.
I think this latter approach is typical of lots of numerical software
written
using conventional arithmetic (e.g. double float). Can one do better
with variable (arbitrary) precision? Yes, sometimes.
RJF