Study On Natural Calculationmethod Based On Population Sizedynamic Control Strategy

At present,most organisms live in groups,and when they reproduce to a certain period,they will follow similar survival rules.Therefore,the research on natural calculation methods reflecting "group behavior" has been a hot topic.At the same time,large populations mean complex and fragmented information,unsatisfactory algorithm performance and less effort to optimize.This paper proposes a method based on population size dynamic control to solve these problems.The main research results of it are as follows:In order to eliminate the limitations of current population grouping methods and combine with the characteristics of particle flight paths,a population grouping adaptive dynamic control strategy is proposed.Based on the Gaussian fitting curve as reference,the strategy is grouped according to monotonicity,and the velocity of the particles crossing the boundary before flight is inverted,so as to ensure the accuracy of the internal information of the subpopulation and to quickly converge to the trough of the current subpopulation.A dynamic population size control method based on Euclidean distance was proposed.The core circle domain was established by introducing Euclidean distance,and the feedback information of the core circle domain was used to dynamically control the population size.Then,based on the change rule of the core circle domain before and after iteration,individuals inside and outside the circle domain were added or deleted.To prevent premature convergence or low efficiency caused by too small or too large population size.The above two control strategies for population size balance the search capability of the algorithm before and after iteration and improve the accuracy of the algorithm.Moreover,these two strategies have nothing to do with each algorithm evolution operation,therefore,it is appropriate to use in the related population optimization algorithm,not far-fetched.These two strategies are combined with particle swarm optimization algorithm,genetic algorithm,differential evolution algorithm,and they are respectively tested with the relevant algorithms in recent years by function,comparing the optimization results and the applicability of the two strategies,highlighting the advantages of the strategy proposed in this paper.