Evolution of structure and learning — A GP approach

Abstract

Recently evolutionary algorithms have been shown to be successful in evolving optimal neural network topologies and also novel learning rules. Genetic programming is a new paradigm that has proved to solve a number of complex problems in various domains. In this paper, I have suggested a novel approach to show how genetic programming can be an effective tool in evolving neural networks that work on the principles of interaction, competition, selforganization and adaptation, that is a self-organizing neural network. Can we evolve new learning algorithms with this approach? Can we extend this approach to evolve complex selforganizing systems? Can we employ this approach to evolve and simulate the mechanisms that are found in various sub-systems in the brain and hence for biological modelling? In this work, I have attempted to answer some of these questions.