Evolution of Gene Coordination Networks

Abstract

A new model for the incorporation of learning with simulated evolution is presented. The model uses gene coordination networks to control gene expression. Alleles at a locus compete for expression by matching up to the network. Reinforcement is achieved through choice dynamics where gene expression will be decided by competing environmental states. The result is a epistasis model containing both plasticity and mean loci. Solutions obtained are adaptive in the sense that any changes in the environment will bring about a spontaneous self-organization in the pattern of gene expression resulting in a solution with (near) equivalent fitness. Additionally the model makes the search for structures through neutral or near neutral mutation possible. The model is tested on two standard job-shop scheduling problems which demonstrate the novelty of the approach.