Interactive genetic algorithm based on extinction mechanism
Yong Zhang · Control theory & applications · 2006
The premature convergence and a user's fatigue are two issues in interactive genetic algorithm. The extinction mechanism is put forward, which will help to decrease the search space and enhance the algorithm's performance. The mechanism makes use of the history evolution information to identify the taboo value to extinguish the inferior species and by preventing duplicating to extinguish the inferior individuals. The mechanism also helps the rules of extinction and survival to hold down each other by means of the search space partition. To validate the proposed mechanism, the formula for the variation of individual numbers of the taboo subspace and valid subspace with the evolutionary generation are deduced, and the parameters sensitivity of the proposed mechanism's performance is also analyzed. Furthermore, the high performance of the proposed mechanism is proved by the fast shrinkage of the valid space and the shorter maximum time to converge. Finally, the efficiency of the proposed approach is shown in the comparison experiments. The results manifests that the shrinkage of search space can significantly avoid the premature and lessen a user's fatigue.