A hybrid genetic algorithm with tabu search and Bayesian optimization

摘要

Genetic Algorithms (GA) are parallel search methods with selection mechanisms for solution candidates and genetic operation mechanisms for selected candidates. Successful GAs are equipped both mechanisms. Among them, in the recent literature, the tabu search is a powerful one for multimodal and/or multiobjective problems, while the Bayesian Optimization Algorithm (BOA) shows very high performance for GA-hard problems. Based on this, the paper proposes a novel hybrid genetic algorithm for multimodal function optimization problems. The basic idea is very simple: we introduce tabu lists for the solution selection process and BOA for genetic operations. Intensive experiments have shown the proposed method overperforms conventional GAs in finding multiple solutions and fast convergence.