Font Size: a A A

Research On Evolutionary Algorithm For Solving Single-objective And Multi-objective Optimization Problems

Posted on:2018-09-04Degree:MasterType:Thesis
Country:ChinaCandidate:L LiuFull Text:PDF
GTID:2348330536483304Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
As a kind of heuristic search algorithm,evolutionary algorithms(EAs)are more and more concerned by researchers because of its fast computing power based on computer simulation without cumbersome mathematical formula deduction.Optimization problems widely exist in science research and engineering experience.Hence,two kinds of improved evolutionary algorithms are proposed for solving single-objective and multi-objective optimization problems,respectively.For single-objective optimization problem,this paper deeply studies the differential evolution algorithm,and proposes an improved algorithm for differential evolution based on interactive information.The traditional differential mutation strategy often ignores the value of the interactive information from population,and randomly selects individual as target vector,direction vector and difference vector.To deal with the limitations,a novel interactive information based scheme is proposed.The interaction between individuals is explored to guide the direction in searching progress for better performance.Experiments shows the effectiveness of INN strategy.That is to say,the improved algorithms enhance the performance of the original algorithms on most test problems.For multi-objective optimization problem,this paper studies the multi-objective optimization algorithm based on decomposition(MOEA/D)and proposes a novel MOEA/D with ensemble operators and neighborhood sizes(MOEA/D-EON).In classical algorithm,different mutation operators and different neighborhood sizes have different effects on the performance.Hence,this paper presents a novel algorithm,called MOEA/D-EON.Four neighborhood sizes and four mutation strategies are combined in pairs as a part of candidate pool.Better combination performance is more likely to be selected.Numerical results indicate that the proposed MOEA/D-EON algorithm significantly outperforms the classic MOEA/D algorithm and its existing advanced MOEA/D variants.
Keywords/Search Tags:evolutionary algorithm, differential evolution, multi-objective evolutionary algorithms based on decomposition, global optimization
PDF Full Text Request
Related items