Multi-Stages Genetic Algorithms: Introducing Temporal Structures to Facilitate Selection of Optimal Evolutionary Paths

Abstract

Standard genetic algorithms (GA) are often confronted with the problem of rapid premature convergence. The loss of diversity in a population usually slows down evolution to a significant extent. In this paper, we explore the use of an original strategy called the multi-stages GA as a means of impeding premature convergence and optimizing evolutionary progresses at the same time. The algorithm introduces the idea of temporally organizing an evolutionary process. Evaluation results show that the multi-stages GA significantly outperforms the standard GA.