2-306 変数間の依存関係の有無に着目した実数値 GA の提案とその性能評価

概要

論文の詳細を見る
The Genetic algorithm (GA) is a powerful optimization framework inspired by the evolution process of natural life. GAs for function optimization can be categorized into two groups : bit-string GAs and real-coded GAs. UNDX+MGG is a real-coded GA that has shown good performance on multi-modal functions with epistasis among parameters, which are difficult to solve for conventional methods. However, the conventional UNDX+MGG has a problem from a viewpoint of search efficiency because the conventional UNDX+MGG always assumes epistasis among the all parameters and searches the all parameters at the same time. In this paper, we propose a new method for making the search effective by estimating epistasis among parameters and reducing the number of parameters to be simultaneously searched.
2002-11-14