Graph-based Clustering for Multi-objective Evolutionary Algorithm

Multi-objective optimization is a challenging task in artificial intelligence and optimization field, as it deals with multiple conflicting objectives to be optimized. This paper suggests an approach for decomposing a multi-objective optimization problem (MOP) into a number of simple multi-objective optimization sub-problems. The proposed method takes the advantage of nearest neighbor algorithm. it first maps the evolutionary population to a kNN graph then using the local traits of population it divides them into a set of smaller fixed size clusters. These clusters become sub-problems. With the use of these sub-problems it solves MOPs in a collaborative way. Each sub-population is a cluster that receives computational effort separately. Experiments have been conducted to compare the proposed method with state-of-the-art multi-objective evolutionary algorithms, results show the effectiveness and superiority of the proposed method compared to the other ones.