領域適応型遺伝的アルゴリズムの開発 : 精度および求解性の向上のためのオペレータの提案

概要

論文の詳細を見る
In design problems, design variables are roughly divided into three types; integer, discrete and continuous. And in practical design, they are composed by mixed variables of these three types of design variables. And these types of the problems are hard to optimize, because there exist no particular method to treat these kinds of problems. Although genetic algorithms have been said to be the powerful tools for mixed optimization, it cannot treat continuous variables by its nature and most of the cases they cannot search global solution right away. In order to overcome these situation, we have been developed the adaptive range genetic algorithms, which will change the searching range adaptively from the information obtained in the previous generation and showed its effectiveness to mixed optimization. However, there are some lack in certainty in finding global optimum and also the accuracy will change in each trial. In this study, we will propose some new operators to raise certainty to obtain global optimum and also to reduce the number of function call. We demonstrate the proposed method to two type of bench mark test and show the effectiveness of the proposed method.
一般社団法人日本機械学会の論文
1999-10-25