粒子群最適化アルゴリズムのパラメータ選択と拡張

概要

論文の詳細を見る
Particle swarm optimization (PSO) is a powerful tool for designing antennas, solving inverse scattering problems, and so on. The algorithm of PSO is controlled with several parameters. Unless the parameters are selected appropriately, the search efficiency of PSO drops significantly. There are, however, no clear rules for the selection, and users have considerable difficulty to use PSO efficiently. This paper proposes a guideline and a new technique “particle refresh” for the selection to make the algorithm easy-to-use and keeping high searching-performance. The hybridization between PSO and conjugate gradient method is also discussed to utilize their complementary advantages in global exploration and local exploitation, where “attraction basin recognition” algorithm is proposed to recognizing the attraction basin area of local minima and help the algorithm to escape from local minima certainly and efficiently.