Alpha evolution: An efficient evolutionary algorithm with evolution path adaptation and matrix generation

Metaheuristics involve information extraction and utilization processes to generate more promising solutions. However, the background of excessive metaphor has led to ambiguity in the computational process. To solve this problem, this paper proposes a novel evolutionary algorithm called alpha evolut...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Engineering applications of artificial intelligence Jg. 137; S. 109202
Hauptverfasser: Gao, Hao, Zhang, Qingke
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.11.2024
Schlagworte:
Alpha operator
Evolutionary computation
Non-metaphor algorithm
Numerical optimization
Real-world applications
Real-world applications
Non-metaphor algorithm
Numerical optimization
Alpha operator
Evolutionary computation
ISSN:0952-1976
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Metaheuristics involve information extraction and utilization processes to generate more promising solutions. However, the background of excessive metaphor has led to ambiguity in the computational process. To solve this problem, this paper proposes a novel evolutionary algorithm called alpha evolution (AE). It updates the solution using the alpha operator with the adaptive base vector and the random and adaptive step sizes. First, sample candidate solutions to construct the evolution matrix. Estimate the population state through diagonal or weighted operations of the evolution matrix. To enhance the correlation of estimates for each generation, two evolution paths accumulate the estimate results and achieve the base vector adaptation. Second, the composite differential operation constructs the adaptive step size to estimate the problem gradient, which is used to accelerate the convergence of AE. Finally, the attenuation factor alpha adaptively adjusts the random step size generated based on search space to balance exploration and exploitation. AE was comprehensively verified regarding its search bias, invariance, scalability, parameter sensitivity, search behavior, qualitative indicators, exploration and exploitation, convergence, statistics, and complexity. In numerical simulation, AE was compared with 106 algorithms on the CEC’17 benchmark announced at the 2017 congress on evolutionary computation (CEC). Furthermore, AE was applied to solve multiple sequence alignment and engineering design problems. The evidence shows that AE is competitive in exploration and exploitation, convergence speed and accuracy, avoiding local optima, applicability, and reliability. The source code of AE is publicly available at https://github.com/tsingke/AlphaEvolution. •Propose a new non-metaphor evolutionary algorithm called alpha evolution.•Propose a base vector adaptation mechanism for global optimization problems.•Mathematical models not based on metaphors are more intuitive and understandable.•A total of 106 optimization algorithms are used as performance comparisons for AE.•Numerical experiments and practical applications evaluate the competitiveness of AE.
ISSN:0952-1976
DOI:10.1016/j.engappai.2024.109202