Runtime analysis of adaptive selection variation operators in evolutionary algorithm with reinforcement learning
摘要
Variation operators have great impact on the performance of evolutionary algorithms (EAs) because they determine the way of EAs constructing the new solutions during the evolutionary process. However, finding efficient variation operator for various kinds of optimization problems is a very challenge task in the field of evolutionary computation. This paper proposes an adaptive variation operator EA with using reinforcement learning. Specifically, the proposed EA adaptively selects variation operator from a set of simple operators to generate the offspring individuals by using Q-learning method in each generation. Theoretical analyses on a set of Pseudo-Boolean functions show that the expected runtime of the proposed algorithm is better than or equivalent to that of the well-studied