摘要:This study investigates Estimation of Distribution Algorithms (EDAs) based Bayesian networks with KS learning method. The EDAs based Bayesian networks are used to analyze the effect of learning the best structure in the search. By using KS learning method that can learn optimal Bayesian networks, two important issues in EDAs are studied. First, we discuss that whether learning a more perfect depending model leads to a better behave of EDAs. Second, when a perfect learning is accomplished, we are able to observe that how is the problem structure transformed into the probabilistic model. Several different kinds of experiments have been conducted. The experimental results show that when the accuracy of the learning is increasing, the quality of the problem information learned by the probabilistic model can also be improved. However, the improvements in model accuracy do not mean a more efficient search at all times.