Multi-objective immune algorithm with dynamic memetic Cauchy mutation

Abstract

In this paper, a novel immune algorithm with dynamic memetic Cauchy mutation (DMCMIA) for multi-objective optimization is proposed. The idea of memetics is incorporated into the mutation process and a dynamic memetic Cauchy mutation (DMCM) operator is developed. The DMCM operator combines global exploration and local refinement efficiently, which adopts a generation-dependent parameter to guarantee a good balance between global search and local search. Comparison is made to another multi-objective optimization algorithm, nondominated neighbor immune algorithm, termed as NNIA, in solving five ZDT and five DTLZ standard test problems. Simulation results based on coverage of two set, convergence metric and spacing show that DMCMIA performs better than NNIA in generating approximations to the true Pareto front. In addition, the effectiveness of the novel dynamic memetic Cauchy mutation is verified by comparison to polynomial mutation and Gaussian mutation, the experimental results reinforce the advantage of the DMCM operator.