Scholarlink

<p>Multi-population techniques have been widely used to improve the optimization performance of nature-inspired optimization algorithms and parallel multi-population method is one of the mainstream methods. However, there is still a lack of sufficient theoretical investigations to provide a theoretical foundation for parallel multi-population method. For a better understanding of the parallel multi-population method, this paper aims to take a first step towards rigorously analyzing the effectiveness and performance of parallel multi-population based multi-objective evolutionary algorithms. Hence, we propose a simple parallel multi-population method based multi-objective evolutionary algorithm, called SPMPEA, with a simple sub-population communication strategy. In <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\varvec{COCZ}_{\varvec{TL}}\)</EquationSource> </InlineEquation> problem, the SPMPEA can find the whole Pareto front with a probability of at least <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\varvec{\Omega ((\frac{1}{n})}^{\varvec{n}^{\varvec{2}}}\varvec{)}\)</EquationSource> </InlineEquation> which is better than GSEMO. The expected runtime of SPMPEA on <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(\varvec{COCZ}\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\varvec{OJZJ}\)</EquationSource> </InlineEquation> problems are <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(\varvec{O}\varvec{(n}^{\varvec{2}} \varvec{\log n)}\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(\varvec{O((n-2k) n}^{\varvec{k}}\varvec{)}\)</EquationSource> </InlineEquation> respectively, which is equal to GSEMO. The result in <InlineEquation ID="IEq7"> <EquationSource Format="TEX">\(\varvec{COCZ}_{\varvec{TL}}\)</EquationSource> </InlineEquation> problem shows that the population diversity introduced by parallel multi-population method provides MOEAs with the capability to solve complex problems, and the results of runtime analysis show that parallel multi-population method cannot enhance the performance of MOEAs without a proper sub-population communication strategy.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A simple parallel multi-population evolutionary algorithm for analysis

  • 期刊论文

  • 发表日期：

    2026年5月14日

查看全文
  • Tianyi Yang,
  • Yuren Zhou

摘要

Multi-population techniques have been widely used to improve the optimization performance of nature-inspired optimization algorithms and parallel multi-population method is one of the mainstream methods. However, there is still a lack of sufficient theoretical investigations to provide a theoretical foundation for parallel multi-population method. For a better understanding of the parallel multi-population method, this paper aims to take a first step towards rigorously analyzing the effectiveness and performance of parallel multi-population based multi-objective evolutionary algorithms. Hence, we propose a simple parallel multi-population method based multi-objective evolutionary algorithm, called SPMPEA, with a simple sub-population communication strategy. In \(\varvec{COCZ}_{\varvec{TL}}\) problem, the SPMPEA can find the whole Pareto front with a probability of at least \(\varvec{\Omega ((\frac{1}{n})}^{\varvec{n}^{\varvec{2}}}\varvec{)}\) which is better than GSEMO. The expected runtime of SPMPEA on \(\varvec{COCZ}\) and \(\varvec{OJZJ}\) problems are \(\varvec{O}\varvec{(n}^{\varvec{2}} \varvec{\log n)}\) and \(\varvec{O((n-2k) n}^{\varvec{k}}\varvec{)}\) respectively, which is equal to GSEMO. The result in \(\varvec{COCZ}_{\varvec{TL}}\) problem shows that the population diversity introduced by parallel multi-population method provides MOEAs with the capability to solve complex problems, and the results of runtime analysis show that parallel multi-population method cannot enhance the performance of MOEAs without a proper sub-population communication strategy.