A Study on Parallel Varying Mutation in Deterministic and Self-Adaptive GAs with 0/1 Multiple Knapsack Problems (特集:進化的計算)

概要

論文の詳細を見る
In this work we study varying mutations applied parallel to crossover in generational deterministic and self-adaptive varying mutation GAs and compare them with the conventional generational model of varying mutations that apply mutation mostly serial to crossover. Experiments are conducted with several subclasses of 0/1 multiple knapsack problems. We found that varying mutation parallel to crossover can be a more effective and efficient framework than the conventional model of varying mutations in both deterministic and self-adaptive Gas to achieve faster convergence velocity and higher convergence reliability. We also found that the conventional model of varying mutations affects negatively the self-adaptive mutation rate control. This strongly suggests that the conventional model of varying mutation GAs may not be appropriate for combining forms of control. Also, we compare deterministic, adaptive, and self-adaptive mutation schedules within the parallel varying mutation model. Best overall performance was achieved by a parallel varying mutation self-adaptive GA.
一般社団法人情報処理学会の論文
2004-02-15