Parallel Memetic Algorithm with Selective Local Search for Large Scale Quadratic Assignment Problems

The extent of the application of local searches in canonical memetic algorithm is typically based on the principle of “more is better”. In the same spirit, the parallel memetic algorithm (PMA) is an important extension of the canonical memetic algorithm which applies local searches to every transitional solutions being considered. For PMA which applies a complete local search, we termed it as PMA-CLS. We show in this paper that instead of a complete local search, the island model PMA with selective application of local search (PMA-SLS) is effective in solving complex combinatorial optimization problems, in particular large-scale quadratic assignment problems (QAPs). A distinct feature of the PMA-SLS to be noted in our study is the sampling size. We make use of a normal distribution scheme to determine the sampling ratio. Empirical study on large scale QAPs with PMA-SLS and PMA-CLS are presented. It is shown that PMA-SLS arrives at solutions that are competitive to the PMA-CLS at significantly lower computation efforts on the diverse large scale QAPs considered. This we concluded is due mainly to the ability of the PMA-SLS to manage a more desirable diversity profile as the search progresses.