Decomposition-based multi-objective evolutionary algorithm with customized evolution strategy according to population state

Multi-objective evolutionary algorithms (MOEAs) are designed to obtain a set of solutions that closely approximate the Pareto front and maintain uniform distribution along it. However, most MOEAs apply the same strategy throughout the whole evolutionary process, which may introduce certain limitatio...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Applied soft computing Jg. 183; S. 113639
Hauptverfasser: Chen, Lexing, Li, Taiyong, Zhu, Donglin, Deng, Wu
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier B.V 01.11.2025
Schlagworte:
Customized emphasis procedures
Decomposition
Multi-objective evolutionary algorithm
State identification procedure
Customized emphasis procedures
Decomposition
State identification procedure
Multi-objective evolutionary algorithm
ISSN:1568-4946
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Multi-objective evolutionary algorithms (MOEAs) are designed to obtain a set of solutions that closely approximate the Pareto front and maintain uniform distribution along it. However, most MOEAs apply the same strategy throughout the whole evolutionary process, which may introduce certain limitations. To address this issue, an MOEA based on decomposition (MOEA/D) where population can adaptively evolve based on its state is proposed to improve the convergence or diversity of solutions at the most appropriate time. First, for robustly identifying the state, a flexible sliding window is used to record the historical convergence information of the population. Then, with the support of a state identification procedure, the convergence emphasize procedure and the diversity emphasize procedure are designed to emphasize the convergence and diversity of solutions respectively. Experimental results on 39 different test problems and 3 real-world problems demonstrate that the proposed algorithm outperforms the compared algorithms including three dominance-based MOEAs, six decomposition-based MOEAs and one indicator-based MOEA. •Population state is automatically identified as either converging or stagnating.•A modified Tchebyche aggregation function is applied for converging state.•Diversity is emphasized by a set of improved operations for stagnating state.•Experimental results validate the performance of the proposed approach.
ISSN:1568-4946
DOI:10.1016/j.asoc.2025.113639