Mutation Matrix in Evolutionary Computation: An Application to Resource Allocation Problem

A new approach to evolutionary computation with mutation only is developed by the introduction of the mutation matrix. The method of construction of the mutation matrix is problem independent and the selection mechanism is achieved implicitly by individualized and locus specific mutation probability based on the information on locus statistics and fitness of the population, and traditional genetic algorithm with selection and mutation can be treated a special case. The mutation matrix is parameter free and adaptive as the mutation probability is time dependent, and captures the accumulated information in the past generations. Three methodologies, mutation by row, mutation by column, and mutation by mixing row and column are introduced and tested on the resource allocation problem of the zero/one knapsack problem, showing high efficiency in speed and high quality of solution compared to other traditional methods.