From the Chromosome to the Neural Network

Abstract

A proposal for a model of morphogenesis process taking inspiration from biology is presented in this paper. This process uses a chromosome as a production system to create an artificial neural network. It starts with a single cell containing the chromosome. Cells can divide and establish connections among them. Both structure and weights of the neural network are defined by the morphogenesis process. An application to a neural network driving an autonomous mobile robot is presented which exhibits encouraging first results. 1 Introduction A promising direction for the search for optimal neural networks could be the use of genetic algorithms in order to evolve the structure and the weights of neural networks. The second section will point out the coding problem and give an overview of the state of the art in this research area. Then, in section three, it will be explained how biological inspiration could help to design a model for a morphogenesis process able to generate a neural networ...