AUTOMATIC DEFINITION OF MODULAR NEURAL NETWORKS

This article illustrates an artificial developmental system that is a computationally efficient technique for the automatic generation of co mplex artificial neural networks (ANNs). The artificial developmental system can develop a graph grammer into a modular ANN made of a combin ation of simpler subnetworks. A genetic algorithm is used to evolve co ded grammars that generate ANNs for controlling six-legged robot locom otion. A mechanism for the automatic definition of neural subnetworks is incorporated. Using this mechanism, the genetic algorithm can autom atically decompose a problem into subproblems, generate a subANN for s olving the subproblem, and instantiate copies of this subANN to build a higher-level ANN that solves the problem. We report some simulation results showing that the same problem cannot be solved if the mechanis m for automatic definition of subnetworks is suppressed. We support ou r argument with pictures that describe the steps of development, how A NN structures are evolved, and how the ANNs compute.