Summary
Radial Basis Function Networks (RBFNs) have shown their capability to be used in classification problems, and therefore many data mining algorithms have been developed to configure RBFNs. These algorithms need to be given a suitable set of parameters for every problem they face, thus methods to automatically search the values of these parameters are required. This paper shows the robustness of a meta-algorithm developed to automatically establish the parameters needed to design RBFNs. Results show that this new method can be effectively used, not only to obtain good models, but also to find a stable set of parameters, available to be used on many different problems.
See the full content of this document
Extract
Study of the Robustness of a Meta-Algorithm for the Estimation of Parameters in Radial Basis Function Neural Networks Design
1. Introduction
Radial Basis Function Networks (RBFNs) are one of the most important Artificial Neural Network paradigms in the machine learning field. They have been successfully used in many areas such as pattern classification, function approximation, and time series prediction, among others. RBFNs have interesting characteristics such as their simple topological structure and the fact that outputs can be easily explained (since they depend on the region of the input space in which the input pattern is located). RBFNs also represent a special kind of nets since, once the structure has been fixed, the optimal set of weights linking hidden to output neurons can be analytically computed. Scientists have applied data mining techniques to the tasks of finding the optimal RBFNs that solves a given problem, as can be see in Section 2.) Thus, many methods have been developed to face this problem, all of them sharing the same disadvantage: they need to be given a good parameter setting in order to work properly. Selecting good, or even the best, possible data mining method and finding a reasonable parameter setting are very time co...See the full content of this document
Sponsored links
