On the significance of the permutation problem in neuroevolution
Stefan Haflidason, R.S. Neville · 2009
In this paper we investigate the impact of the Permutation Problem on a standard Genetic Algorithm evolving neural networks for a difficult control problem. Through the use of Price's equation and an explicit enumeration of permutations within the population we demonstrate that for the given problem and representation the Permutation Problem is not as serious a concern as previously thought. In addition we present the concept of incompatible representations as a more useful guide for what to avoid in the evolution of neural networks.