A Multi-Objective Genetic Algorithm Based on Density

摘要

This paper presents a new kind of MOEA, namely DMOGA (Density based Multi-Objective Genetic Algorithm). After discussing the influence function and the density function, we employ density of a solution point as its fitness in order to make the DMOGA perform well on diversity. And then, we extend our discussions to fitness assignment and computation, pruning procedure when the non-dominated set is bigger than the size of evolutionary population, and selection from the environmental selection population. To make DMOGA more efficient, we propose to construct the non-dominated set with the Dealer's Principle. We compare our DMOGA with two popular MOEAs, and the experimental results are satisfactory.