20112 潜在的に局所的な探索の概念と集中化機能の評価への応用(OS11 設計・システム)

概要

論文の詳細を見る
A concept of potentially local search is introduced and used for the evaluation of intensification performance in local search optimization process. The search space is divided by the basin of locally optimal solution and the search dynamics is observed as a transition process from basin to basin. Numerical experiment for the Metropolis algorithm is performed on random Euclidean traveling salesman problems. The result shows that unidirectional transition to better basin appears at low temperatures, which indicates that the present search dynamics has a good intensification characteristic there. The optimum temperature for potentially local search seems to be determined depending on the trade-off between the quickness of the intensification effect and the goodness of the reachable basin. The increase in the problem size seems to widen the difference in optimization characteristics between the real and the potential search dynamics.
一般社団法人日本機械学会の論文
2005-03-17