Font Size: a A A

Multi-objective Optimization Algorithm For Social Learning

Posted on:2019-09-24Degree:MasterType:Thesis
Country:ChinaCandidate:J R WuFull Text:PDF
GTID:2428330566976287Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The multi-objective optimization problem is widely applied in various scientific fields and engineering projects.The traditional multi-objective optimization algorithm mostly transforms multiple targets into multiple single targets by adding weight vectors,and then optimizes the single objective.However,the value of weight vector itself is an optimization problem.With the development of evolutionary optimization algorithm in recent years,evolutionary optimization algorithm has been used to solve multi-objective optimization problems more and more.However,with the complexity of the problems,more and more factors are considered,resulting in more and more optimization variables,thus getting better Pareto frontiers is becoming more and more difficult.So far,although there are many different methods of solving multi-objective problems,it is not much used in the multi-objective problem of large scale decision space.On the other hand,as a variety of particle swarm optimization,the social group particle swarm optimization(PSO)algorithm has obtained a better solution because of its diversity in solving the problem of large-scale single objective optimization.Therefore,in this paper,we first propose the use of social population particle swarm optimization algorithm combined with decomposition strategy to solve large-scale multi-objective optimization problems.On the other hand,in order to reduce the number of evaluation of optimization problems and save the cost of computing,we also propose a support vector machine assisted social learning particle swarm optimization(SVM),which trains support vector machines by using the location of history and the corresponding adaptive values,and determines whether the new species may be non dominant.For non dominated solutions,the real objective function is calculated.The results of simulation experiments on different test functions show that the two methods proposed in this paper can get a better set of solutions than the MOEA/D algorithm.
Keywords/Search Tags:Multi-objective optimization algorithm, Particle swarm optimization, Decomposition strategy, Neighbor social learning, Support vector machine
PDF Full Text Request
Related items