Font Size: a A A

Research Of Differential Evolution With New Mutation And Crossover Operators

Posted on:2017-10-05Degree:MasterType:Thesis
Country:ChinaCandidate:S X ZhangFull Text:PDF
GTID:2348330503967134Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Evolutionary algorithm which was inspired from natural species evolution has become the research focus in computing science and there were many branches emerged. Over the past few years, different evolution(DE) had raised more and more researchers' interest and has been one of the most competitive evolutionary algorithms. Three genetic operators, namely, mutation, crossover and selection are involved in DE. Among them, mutation and crossover operators are the main research topics.To remedy the drawbacks of the current DE operator research, this paper proposes the elitism centroid based mutation and crossover operators and the linear combination based mutation and crossover operators. The main contributions of this dissertation are as follows:By analyzing the stagnation issue which hinders the performance improvement of DE, this paper presents a CMX framework based on the proposed elitism centroid based mutation(CM) and elitism centroid based crossover(CX) operators. The framework continuously evaluates the state of the population individual and once stagnation is detected, CM and CX operators instead of the classical mutation and crossover operators are performed to help the algorithm converge. Experiments conducted on the standard test function suit shown that the introduced framework is capable of handing the stagnation problem and enhancing the performance of DE. Moreover, comparison between the presented framework and the successful-parent-selecting(SPS) framework in literature indicated that CMX framework is more efficient in dealing with stagnation.The current DE operator research prefers to utilize the information from single individuals in the population, but neglects the guiding information from several promising individual. With this consideration, the paper proposes a linear combination based mutation(LCM) and a linear combination based crossover(LCX) operator and further implements a new DE variant LCDE. Numerical results confirmed the enhanced searching capability of the proposed LCM operator compared to other competitive mutation operators in literature, such as ?current-to-pbest? and ?current-to-gr_best? and the superiority of LCDE over other state-of-the-art DE variants.
Keywords/Search Tags:Differential evolution, mutation and crossover operator, differential evolution variant, global optimization
PDF Full Text Request
Related items