領域適応型Differential Evolutionの提案

概要

論文の詳細を見る
This paper proposes a new method called Adaptive Range Differential Evolution (ARDE). The basic idea can be obtained from the Adaptive Range Particle Swarm Optimization (ARPSO). In the ARDE, the active search domain range consists of the mean and the standard deviation of each design variable. Also the best position is included in this search range. Therefore, the active search domain range is newly defined by utilizing the mean and the standard deviation of each design variable, and the best position. The detailed procedure of the new active search domain range is described in this paper. To keep the best position for the new active search domain range will lead to the wide search range. Of course, the newly active search domain range will shrink through the search iteration. As the result, a highly accurate global minimum can be found. The effectiveness and validity of the ARDE are examined through typical benchmark problems. It could be found through benchmark problems that the ARDE can find a global minimum with the small number of function evaluations in comparison with basic Differential Evolution. ©2013 The Japan Society of Mechanical Engineers.
日本機械学会 = The Japan Society of Mechanical Engineers, JSMEの論文