Summary
International Symposium on Nonlinear Theory and its Applications
2005
Session Number:1-3-3
Session:
Number:1-3-3-5
A Simple Radial Basis ART Network: Basic Learning Characteristics and Application to Area Measurement
Masahide Ohki, Torikai Hiroyuki, Toshimichi Saito,
pp.262-265
Publication Date:2005/10/18
Online ISSN:2188-5079
DOI:10.34385/proc.40.1-3-3-5
PDF download (195.8KB)
Summary:
A simple radial basis adaptive resonance theory network (RBART) is proposed. Based on unsupervised learning, the RBART can approximate a set of input data by a set of circle-shaped categories. We clarify some generalities of learning characteristics, e.g., number of created categories and their total area are almost independ on distribution of inputs but depend on size of the input data set. We then propose that the RBART can be applied to robust measurement of area of given figures.