Re: how to evaluate the addition of millions of functions dynamically and efficiently?

From: jim green (jjg_at_withheld.org.de)
Date: 08/16/04 

Date: Mon, 16 Aug 2004 12:25:48 GMT

"networm" <networm8848@yahoo.com> writes:

> I am also interested in the "RBF monopole" you've mentioned in your
> previous post.
>
> Do you think it will give me a drastic performance improvement? If so, I am
> going to dive into it and learn how to make my program a lot faster...

I have never implemented it, but the idea is rather similar to the
FFT - break the problem into smaller simpler subproblems, and recurse.

I understand that the implemetation is rather involved, so I'd suggest
that you consider using it only if
- your centres are genuinely scattered,
- you basis function does not have compact support.
Otherwise I would guess that a local search would be quicker and a lot
easier to implement.

As I say, I'm no expert on this, others may correct me.

There are quite a few papers in the (mathematics) literature, I'd
recommend that you look at those before proceeding.

-j

Relevant Pages