Relationship between success ratios of hybrid algorithms and those of basic algorithms on bimodal MaxSAT problems
Xinsheng Lai, Hong Hu, Qing Yan, Yulin Zhou · 2016
Empirical results showed that it is a promising way to design more powerful algorithms by combining different algorithms. It is interesting to investigate the relationship between the success ratio of a hybrid algorithm and those of basic algorithms on which the hybrid algorithm is based. In this paper, we analyze the relationship between the success ratios of two hybrid algorithms and those of basic algorithms on bimodal MaxSAT problems. Both hybrid algorithms combine a hill-climbing RandomWalk algorithm with a local (1+1) evolutionary algorithm (EA) in different ways. The success ratio refers to the percentage of successful runs. A successful run means that an algorithm finds a globally optimal solution when a termination condition has been reached. Analysis shows that on bimodal MaxSAT problems neither hybrid algorithms improves the success ratio with respect to those of the hill-climbing RandomWalk algorithm and the local (1+1) EA, which is tested by experiments.