From: spellucci@mathematik.tu-darmstadt.de (Peter Spellucci)
Subject: Re: multidimensional fit
Date: 7 May 1999 12:50:14 GMT
Newsgroups: sci.math.num-analysis

In article <7gsbir$lho$1@cnn.Princeton.EDU>,
 androula@titan.Princeton.EDU (Ioannis Androulakis) writes:
|> I am looking for any public domain software that
|> determines approximations (polynomial or else)
|> of maps of the from f:x->y, where x is an n-dimensional
|> (n>2) input vector and y scalar.
|> Please forward any suggestions to ipandro@erenj.com
this question is much too unprecise to be answered usefully. the answer
depends very much on the kind of data you have concerning "f".
grid data in R^n? how large is n? how much data? 
scattered data?  can f be computed for arbitrary x and you simply want a cheap
approximation for it? 
possible solutions : multidimensional fft (software available up to n=4)
multidimensional tensorproduct splines (software available up to n=2 )
multidimensional lagrangian interpolation: see new work of thomas sauer 
(advances in computational mathematics 3, 1995, 219--238 and Math. Comp. 64,
1995, 1147--1170)
scattered data fits (software in netlib/toms)
multidimensional least squares fits (software in 
http://plato.la.asu.edu/guide.html)
and and ...
hope this helps
peter