Structured learning in recurrent neural network using genetic algorithm with internal copy operator

Abstract

We compose a genetic algorithm that uses an internal copy operator for recurrent neural network learning. The internal copy operator copies one part of a gene to another part of the same gene. We show that the proposed algorithm accelerates learning. We also show that the internal copy operator organizes the structure in the network. The organized structure improves the learning ability and makes it possible to acquire a set of limit cycles easily.