A Centroid Guided Cluster Transformation for Dynamic Multi-Objective Optimization Algorithm

In recent years, prediction-based algorithms have made significant progress in solving dynamic multi-objective optimization problems (DMOPs). However, most existing algorithms only consider information from several consecutive environments and ignore previous search experiences. This article uses th...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2025 IEEE Congress on Evolutionary Computation (CEC) S. 1 - 8
Hauptverfasser: Zeng, Yi, Xia, Xuewen, Lin, Fenglin, Zhang, Yuehui, Liu, Meitong
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 08.06.2025
Schlagworte:
Benchmark testing
cluster transformation
Clustering algorithms
Dynamic multi-objective optimization algorithm
Heuristic algorithms
Long short term memory
Long Short-Term Memory (LSTM)
Optimization
Prediction algorithms
Predictive models
Reliability
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In recent years, prediction-based algorithms have made significant progress in solving dynamic multi-objective optimization problems (DMOPs). However, most existing algorithms only consider information from several consecutive environments and ignore previous search experiences. This article uses the search experience of the Regularity Model-Based Multiobjective Estimation of Distribution Algorithm (RM-MEDA) to propose a dynamic multiobjective optimization algorithm based on the centroid-guided cluster transformation (CGCT-RM- MEDA). When the environment changes, the information from the statically optimized model in the previous environment is used to estimate a new distribution model through a cluster transformation. At the same time, the location of the population distribution is predicted using a Long Short-Term Memory (LSTM) network, which is used to estimate the cluster centers of the model distribution. The empirical study evaluated the performance of CGCT-RM-MEDA using 14 benchmark functions and one performance metric. The experimental results show that CGCT-RM-MEDA outperforms seven peer algorithms in performance.
DOI:10.1109/CEC65147.2025.11043010