Research On Mixed Social Learning Algorithm For Large Scale Optimization Problems

All aspects of people's lives are accompanied by Global optimization problems,and many industrial problems can be transformed into global optimization problems.Along with the rapid growth of processing dimensions,the problem of solving problems is becoming more and more complicated,and the traditional small-scale optimization problem can not meet the needs of the current society.In dealing with large-scale optimization problems,conventional optimization algorithms used to solve traditional global optimization problems are difficult to demonstrate their superiority.A number of new methods and strategies for solving high-dimensional optimization problems have emerged.In this paper,based on the social particle swarm learning algorithm,a hybrid social learning algorithm is proposed for high-dimensional optimization problems.The research in this paper mainly includes the following two aspects:First,the advantage of social learning is that it can learn the merits of others with little cost,to make up for its own shortcomings.Particles in different evolutionary states have different development potentials.The advantage of horizontal learning is to treat different levels of particles differently and to maximize their potential.In this paper,the social learning mechanism and the horizontal learning mechanism are introduced into the particle swarm optimization algorithm(Hybrid Social Learning Particle Swarm Optimization,PSO).Different from the traditional variant of particle swarm optimization is that to guide the learning of particles,the traditional global optimal solution and local optimal solution are discarded.Instead,the particles learn socially from dominant particles in different horizontal intervals.The experimental results show that the algorithm is effective in solving large-scale optimization problems.Secondly,based on the hybrid social learning particle swarm optimization algorithm,an inverse hybrid social learning particle swarm optimization algorithm based on valuation model is proposed(Valuation Model Reverse Hybrid Social Learning Particle Swarm Optimization,VM-HSL-PSO).The hybrid social learning particle swarm optimization algorithm wastes a large number ofevaluation times when solving the horizontal learning interval.By incorporating the valuation model,the hybrid social learning algorithm can continue to find the global optimal solution under the same number of evaluation times.In the process of learning the particle swarm optimization algorithm of mixed society,it is found that the hybrid social learning particle swarm algorithm has a large number of assimilation phenomena in the late stage of evolution.Assimilation makes the particles fall into the learning bottleneck to a large extent and stop at the local optimal solution,losing a wider search area.In view of the above problems,this paper introduces a reverse learning mechanism.when the assimilation phenomenon occurs in the later stage,the particles can go through the reverse social learning to the particles that are worse than themselves,so that to some extent,it can jump out of the local optimal environment and enter a wider area for optimization.Finally,through a large number of data experiments,the results show that the VM-HSL-PSO algorithm has strong optimization ability and has a good optimization effect on large-scale optimization problems.