Research On Information Interaction Mechanism Of Particle Swarm Optimization Algorithm In Large-scale Environmen

Nowadays,large-scale optimization problems widely exist in various fields of the real world.Due to the increasing number of related variables in the optimization problem,the solution space of the optimization problem becomes increasingly complex,which takes a great challenge to the performance of Particle Swarm Optimization(PSO)algorithm.In PSO,the information interaction between the particles is one of the main factors that affect the performance of the particle swarm optimization algorithms.In order to improve the efficiency of the information interaction between particles,this article mainly conducts research on the information interaction of the particle swarm optimization algorithms in large-scale environments.Specifically:①In order to alleviate the parameter sensitivity in large-scale PSO,this paper designs two simple and effective adaptive parameter adjustment strategies.Firstly,a new aggregation indicator is defined based on the difference between the global fitness value and the average fitness value of the swarm.Based on the changes of the indicator during the evolution process,the values of the key parameters in the algorithm are adaptively adjusted to improve the performance of the algorithm;②According to human observation learning theory,a random elite ensemble learning largescale particle swarm optimization algorithm(REELSO)was designed;Firstly,based on the fitness values of the particles in the population,REELSO divides the particles in the current swarm into elite group and non-elite group;Secondly,each non-elite particle in the non-elite group randomly select several individuals from the elite group as the neighbors;Subsequently,the non-elite particle learn from the best individual of the neighbors to complete the cognitive learning process,while the nonelite particle learn from the ensemble of all the neighbors to complete the collective learning process;Finally,this paper further designs an adaptive partition strategy,in which the algorithm adaptively divides the swarm into two groups during the evolution process,enabling the swarm to gradually transform from exploring a huge problem solution space to developing the discovered optimal area,thus avoiding serious loss of diversity;③By investigating the information interaction of the particles in the swarm,this paper designs a random contrastive interaction for particle swarm optimization(RCI-PSO)to deal large-scale optimization problems.Different from the existing low-dimensional particle information interaction methods,for each particle to be updated,RCI-PSO randomly selects several individuals from the current population that are different from each other to construct a random information interaction topology;Subsequently,within the topology,RCI-PSO only selects the two individuals with better fitness values than the particle and the largest difference in fitness values to guide the evolution of the particle.In addition,in order to alleviate the sensitivity of parameters,this paper further designs a dynamic topology adjustment strategy.Finally,this paper has conducted sufficient experiments on two mainstream CEC2010 and CEC2013 large-scale optimization problem test sets.The experimental results show that the three algorithms proposed in this paper could effectively optimize large-scale optimization problems,and achieve better optimization performance compared to existing large-scale optimization algorithms.