Multi-Population Genetic Algorithm Based on Adaptive Learning Mechanism

TL;DR: Wang et al. as mentioned in this paper improved the learning mechanism by adaptively changing the related control parameters, and dynamically controlling the process of learning mechanism, which has great improvement in many aspects of the global optimization, such as convergence speed, the accuracy of the solution, and stability.

Abstract: Traditional genetic algorithm has some disadvantages, such as slow convergence, unstable, and easy to fall into local extreme. In order to overcome these disadvantages, an improved genetic algorithm is proposed in the present study. First, based on the analysis of advantages and disadvantages of learning mechanisms in literature, new improvements of learning mechanisms under the multi-population parallel GA are made. In previous studies, gene patterns from which other individuals can learn will be extracted from the excellent individuals of the population, this study improved the learning mechanism by adaptively changing the related control parameters, and dynamically controlling the process of the learning mechanism. Simulation results show that the new algorithm has a great improvement in many aspects of the global optimization, such as convergence speed, the accuracy of the solution, and stability.