[EM] High precision Yee diagrams

[Warren Smith]

On 1/27/11, Kristofer Munsterhjelm <km-elmet at broadpark.no> wrote:
> Raph Frank wrote:
>> Warren has a way of calculating the score for each candidate using the
>> fast fourier transform.
>
> And what is that way? It would be interesting to know.

--1. The idea is you can use the fourier convolution theorem to
compute the convolution of (say) a Gaussian with a vote-distribution
to find vote totals.  This will run in O(NlogN) flops as opposed to
the naive method taking order N^2 flops, to compute this at all N
pixels.

--2. You (KM) posted some ideas about expressing Yee diagram pixels
via integrals, which is fine except it is not necessarily so easy to
do the integrals.   You might want to look at the "Hermite quadrature
rules" and the book by A.H.Stroud "Approximate calculation of multiple
integrals."



-- 
Warren D. Smith
http://RangeVoting.org  <-- add your endorsement (by clicking
"endorse" as 1st step)
and
math.temple.edu/~wds/homepage/works.html