Genetic algorithm with stochastic automata-controlled, relevant gene-specific mutation probabilities

It has been reported that after prolonged starvation, bacterial cells increase the frequency of mutation and produce new phenotypes advantageous for surviving. This result inspired us to make improvements to genetic algorithms applied for optimum search. Stochastic automata are used to learn the locus of effective genes on the chromosomes of which mutation yields a higher value of the evaluation function. A state probability vector for each automaton generates a mutation probability for the genes at the corresponding locus. This procedure helps the algorithm to escape from being trapped in local maxima or minina. It is shown by simulation studies that the algorithm proposed here is more effective for searching the maximum of multiple peak variable separable functions