基于贪婪更新和自适应扰动的自然计算策略

A natural computing strategy based on greedy renewal and adaptive disturbance

自然计算因其参数少好实现等优点而被广泛应用,但是这些算法存在着寻优精度较低和易陷入局部最优解的问题,为更有效的提高算法性能,提出一种基于贪婪更新和自适应扰动的自然计算策略(Greedy-renewal and Self-adaption Disturb,简称GSD),该方法具备普适性,适用于所有启发式寻优算法。首先使用Faure序列化初始化种群,提高初始解的质量,其次在个体位置更新后,比对更新前后个体的适应度值,如果适应度提高则保留此次更新操作,反之放弃此次更新,最后设计自适应扰动因子,随迭代次数增加调整对个体的扰动几率,提高算法后期的种群多样性。将该策略应用于粒子群算法和遗传算法中,利用经典测试函数验证性能。实验结果表明,贪婪更新和自适应扰动策略改进的算法较其他对比算法表现出了更好的寻优能力,具有普适性。

Natural computing is widely used because of its advantages such as fewparameters and easy implementation,but these algorithms have problems of lowoptimization accuracy and easy to fall into the local optimal solution.In order to improve the algorithmperformancemore effectively,This paper proposes a natural computing strategy based on Greedy renewal and Self-adaption Disturb(GSD),which is universal and suitable for all heuristic optimization algorithms.Firstly,Faure serialization was used to initialize the population to improve the quality of the initial solution.Secondly,after the update of the individual position,the fitness values of the individuals before and after the update were compared.If the fitness was improved,the update operation would be retained;otherwise,the update would be abandoned.Improve the population diversity of the late algorithm.This strategy is applied to particle swarmoptimization and genetic algorithm,and the performance is verified by classical test function.The experimental results showthat the greedy update algorithmand the adaptive perturbation algorithmhave better optimization ability than other comparison algorithms and have universality.