Research On Large-scale Optimization Algorithm Based On Improved Particle Swarm Learning Strategy

Particle Swarm Optimization(PSO)is an important method to resolve complicated problems for its simple principle and easy implementation.However,it can only be applied to dealing with low-dimensional problems.For those high-dimensional(large-scale)optimizations with plenty of variables in engineering applications and scientific researches,the urgent problem is to effectively manage the variables.The PSO effect will deteriorate as the increasing of dimensions,which is so called“the curse of dimensions”.The phenomenon is caused by both of the augment of dimensions and the poor diversity of PSO.This paper put forward two solutions to alleviate the phenomenon of“the curse of dimensions”in PSO's dealing with high-dimensional problems.First,Cooperatively Coevolving Competition Swarm Optimizer(~3)is put forward to dealing with high-dimensional problems.The algorithm is based on the idea of co-evolution to decompose the original large-scale(high-dimensional)optimization problem into multiple lower-dimensional sub-optimization problems.Then,each sub-problem can be resolved individually through the improved competitive optimization algorithm.After testing the optimization results on two common large-scale optimization problem test function sets(CEC2010,CEC2013),we can find that~3can not only get better optimization results but also have advantages in optimizing time.Second,Two-phase Learning-based Swarm Optimizer(TPLSO)is proposed for the problem of poor diversity of PSO.The algorithm is inspired by two factors.On the one hand,the way of group learning in human society,which we call it as Mass learning strategies.It is a strategy that can improve the diversity of particle learning.And Elite learning,on the other hand,influenced by elite groups in social groups,can accelerate the convergence of particles.TPLSO speeds up the convergence of particles while ensuring the diversity of algorithms,which is an important detection standard for evolutionary algorithms.The experimental results on two common large-scale optimization problem test function sets(CEC2010,CEC2013)show that TPLSO has better optimization results than other optimization algorithms with good performance.