forum.alglib.net
http://forum.alglib.net/

How to set basis function of a rbfmodel
http://forum.alglib.net/viewtopic.php?f=2&t=793
Page 1 of 1

Author:  heltonbiker [ Tue Mar 12, 2013 1:16 pm ]
Post subject:  How to set basis function of a rbfmodel

I am trying to use AlgLib's RBF models to replace SciPy in an application.

In Scipy, I use "Thin-Plate" function (
Code:
r**2 * log(r)
), since it gives by far the best-behaved results for my datasets.

I have read the Introduction to RBF's article, and it explicitly mentions the Thin-Plate in the following paragraph:

Quote:
(...) We can get different models with different properties by choosing different basis functions, polynomial terms and smoothing.(...)

(...)
Thin Plate Splines – f(r)=r 2·ln(r) is used as basis function, nonlinearity penalty, polynomial term

All these algorithms, albeit known under different names, are just special cases of same general idea - representation of the model function as linear combination of the radially-symmetric basis functions f(r).


So what I need to know is: "How could I set the basis function of a model to be of the Thin Plate form?".

I searched around the docs, but couldn't figure it out. Thanks for any help!

Page 1 of 1 All times are UTC
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/