Crowding clustering genetic algorithm for multimodal function optimization

Interest in multimodal function optimization is expanding rapidly since real-world optimization problems often require location of multiple optima in a search space. In this paper, we propose a novel genetic algorithm which combines crowding and clustering for multimodal function optimization, and analyze convergence properties of the algorithm. The crowding clustering genetic algorithm employs standard crowding strategy to form multiple niches and clustering operation to eliminate genetic drift. Numerical experiments on standard test functions indicate that crowding clustering genetic algorithm is superior to both standard crowding and deterministic crowding in quantity, quality and precision of multi-optimum search. The proposed algorithm is applied to the practical optimal design of varied-line-spacing holographic grating and achieves satisfactory results.