Abstract
A novel approach to topology and weight evolving artifitial neural networks (TWEANNs) is presented. This comes from the following two considerations. (1) Artificial evolution is designed without recombination which mostly generates an offspring of which fitness value is considerably reduced than parents. Instead of recombination, topological mutations are provided as genetic oprtations for inserting a neuron or for developing a synaptic connection, which are designed carefully to retain the current evaluation value. (2) A new encoding technique such that a string is defined as a set of substrings called operons is introduced. This is based on our idea that each potential function should be encoded into a different part of genetic information. In this research, we examine the evolution behavior by the change in the parameters for yielding good strategies.
