Evolutionary cellular configurations for designing feed-forward neural networks architectures

Abstract

In the recent years, the interest to develop automatic methods to determine appropriate architectures of feed-forward neural networks has increased. Most of the methods are based on evolutionary computation paradigms. Some of the designed methods are based on direct representations of the parameters of the network. These representations do not allow scalability, so to represent large architectures, very large structures are required. An alternative more interesting are the indirect schemes. They codify a compact representation of the neural network. In this work, an indirect constructive encoding scheme is presented. This scheme is based on cellular automata representations in order to increase the scalability of the method.