Guided Pareto Local Search and its Application to the 0/1 Multi-objective Knapsack Problems

Pareto Local Search (PLS) is a generalization of the local search algorithms to handle more than one objective. In this paper, two variants of PLS are examined on the multiobjective 0/1 knapsack problems, compared with three well-known multiobjective EA algorithms, namely SPEA, SPEA2 and NSGA2. Furthermore, A Guided Local Search (GLS) based multiobjective optimization algorithm is proposed, the Guided Pareto Local Search (GPLS). GPLS shows the ability of GLS to set on top of PLS not only to help PLS to escape Pareto local optimal set, but also to enhance its convergence toward and spread over the true Pareto front. Experimental results have shown that PLS can produce results with a very good quality, and proven the effectiveness of the GPLS.