Research on the scope of capacity for different EDAs

Caichang Ding^1,2,3, Wenxiu Peng²

¹State Key Lab of Software Engineering, Wuhan University, Wuhan 430072, China
²School of Computer, Wuhan University, Wuhan 430072, China
³School of Computer Science, Yangtze University, Jingzhou 434023, China

In this paper we investigation the scope of capacity for different EDAs to successfully solve problems, which concern to the mutual effects among the variables. More specifically, we study the learning restrictions that different EDAs confront to solve problems, which can be expressed by some ADFs. The research is conducted in the worst situation. The sub-functions in the ADFs are the same deceptive functions. We think that the capacity for this kind of algorithm are primarily influenced by the probabilistic model they depend on. We employ three different kind of EDAs so as to investigate the effect that the complexity of the probabilistic model has on the behave of the algorithm. Because the population size is crucial for EDAs, we use different population sizes in the experiments. Nevertheless, the results indicate that, in general, enlarge population size is not useful to solve more complex problems.