Accelerated genetic algorithms with Markov Chains

Based on the mutation matrix formalism and past statistics of genetic algorithm, a Markov Chain transition probability matrix is introduced to provide a guided search for complex problem optimization. The important input for this guided search is the ranking scheme of the chromosomes. It is found that the effect of mutation using the transition matrix yields faster convergence as well as overall higher fitness in the search for optimal solutions for the 0-1 Knapsack problem, when compared with the mutation-only-genetic-algorithm,which include the traditional genetic algorithm as a special case. The accelerated genetic algorithm with Markov Chain provides a theoretical basis for further mathematical analysis of evolutionary computation, specifically in the context of adaptive parameter control. © 2010 Springer-Verlag Berlin Heidelberg