Diversity Oriented Local Search for Multi-criteria Optimization

We develop a generic tool for approximating the Pareto front of multicriteria optimization problems using stochastic local search algorithms. Our algorithmic scheme handles problems that the multi-criteria context introduces into the local search framework such as the non-uniqueness of the best neighbor and the potentially large size of the Pareto front. We demonstrate the performance of our algorithm under different configurations and parameters on multi-criteria variants of the quadratic assignment and 0-1 knapsack problems. The adaptation of our scheme to new problems involves a minimal investment.