Differential Evolution Inspired Clone Immune Multi-objective Optimization Algorithm

Abstract

This paper proposes a novel multi-objective optimization algorithm: differential evolution inspired clone immune multi-objective optimization algorithm (DECIMO). The novel algorithm uses a space-filling experimental design named symmetric Latin hypercube design (SLHD) to initialize the population which can obviously improve the uniformity of the individual distribution. A permutation of population individual indexes is generated and then a neighborhood for each population individual is defined according to the permutation. A differential evolution inspired neighborhood recombination operator, which based on the neighbors of each population member, is proposed to balance the exploration and exploitation abilities of the algorithm with no compromise of efficiency. The DE inspired operator is then invoked into the clone immune algorithm (CIA) to solve multi-objective problems (MOPs). We compare the proposed algorithm with NSGA2 and SPEA2 by executing it to 5 famous test functions. The results show that the proposed algorithm can fast converge to the global Pareto front and also can sustain a very uniform distribution. It is a potential algorithm for solving MOPs.