Exploration efficiency of GAS largely depends on parameter values. But, it is hard to manually adjust these values. To cope with this problem, several adaptive GAS which automatically adjust parameters have been proposed. However, most of the existing adaptive GAS can adapt only a few parameters at the same time. Although several adaptive GAS can adapt multiple parameters simultaneously, these algorithms require extremely large computation costs. In this paper, we propose Self Adaptive Island GA(SA1GA) which adapts four parameter values simultaneously while finding a solution to a prohlem. SAIGA is a kind of island CA, and it adapts parameter values using a similar mechanism to meta-CA. Throughout our evaluation experiments, we confirmed that our algorithm outperforms a simple CA using De Jong's rational parameters, and has performance close to a simple CA using manually tuned parameter values.
SUMMARYThe exploration efficiency of GAs depends on parameter values such as the mutation rate and crossover rate. To save the labor of manually adjusting these values, GAs which automatically adjust parameters (adaptive GAs) have been proposed. However, most of the existing adaptive GAs can adjust only a few parameters simultaneously. Although several adaptive GAs can adjust many parameters simultaneously, these algorithms have a large computational cost.In this paper, we propose the Self-Adaptive Island GA (SAIGA) and its asynchronous implementation Asynchronous SAIGA (A-SAIGA). These two GAs are combinations of Meta GA and Island GA, and can adapt many parameters simultaneously with a computational cost equivalent to that of the simple GA. A-SAIGA improves exploration speed by avoiding synchronization between islands.Throughout our evaluation experiments, we confirmed that the performance of these GAs is close to that of the simple GA with optimal parameters. We also confirmed that A-SAIGA outperforms SAIGA in exploration speed.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.