An Autonomous Learning Algorithm of Resource Allocating Network

Abstract

Selecting proper parameters of RBF networks has been a puzzling problem even for batch learning. The parameter selection is usually carried out by an external supervisor. To exclude the intervention by an external supervisor from the parameter selection, we propose a new learning scheme called Autonomous Learning algorithm for Resource Allocating Network (AL-RAN). AL-RAN is an incremental learning algorithm which consists of the following functions: automated data normalization and automated adjustment of RBF widths. In the experiments, we evaluate AL-RAN using nine benchmark datasets in terms of the decision accuracy of data normalization and the final classification accuracy. The experimental results demonstrate that the above two functions in AL-RAN work well and the final classification accuracy of AL-RAN is almost the same as that of a non-autonomous model whose parameters are manually tuned by an external supervisor.

The authors would like to thank Professor Shigeo Abe for his helpful comments and discussions. This research was partially supported by the Ministry of Education, Science, Sports and Culture, Grant-in-Aid for Scientific Research (C) 205002205.