An adaptive memetic Particle Swarm Optimization algorithm for finding large-scale Latin hypercube designs

The Latin Hypercube (LH) design problem arises in computer simulations and is employed for examination and simulation of many physical events. In order to find a (near) optimal LH design, this paper proposes a new version of Particle Swarm Optimization (PSO) algorithm, which uses a population-based optimizer as the evolutionary part and a multiple-local-search procedure as the refinement part of the algorithm. To manage the problem constraints, the proposed algorithm utilizes a Ranked Ordered Value (ROV) rule, which converts the continuous space of solutions to the point-permutation space. Furthermore, to maintain the population diversity, the meta-Lamarckian learning strategy is applied to the local search procedure of the algorithm. In order to test the proposed algorithm, we compare it with a set of existing algorithms on several problem instances. To perform a fair comparison, the best parameters for all the algorithms are found and the experiments are performed based on these parameters. The experimental results show that the proposed algorithm outperforms the existing algorithms in solving large-scale LH design problem.