[comp.ai.neural-nets] Gaussian Radial Basis Functions?

gerald@wiliki.eng.hawaii.edu (Gerald Iseri) (04/22/91)

I've been working on Gaussian Radial Basis
Networks and have recently been thinking 
about the following two questions:
First question:  What are
the advantages of using this type of basis
functions in comparisons to more "traditional"
basis functions such as sinc functions, sinusoids,
polynomial or other orthogonal basis functions?

Second question:  Why use a nonorthogonal basis such
as Gaussians?.  Are there any theoretical results
indicating why?


I have read the papers by Poggio and Girosi
and still have not been able to come to a 
conclusion that justifies the use of 
Gaussian Radial Basis Functions over 
say sinc functions (Shannon's Thm).


--gerald
email replies to: gerald@wiliki.eng.hawaii.edu