Metaheuristics exposed: Unmasking the design pitfalls of arithmetic optimization algorithm in benchmarking

This work unveils design flaws within most metaheuristics, with a specific focus on issues associated with the arithmetic optimization algorithm (AOA). Despite being a simple metaheuristic optimizer inspired by mathematical operations, AOA holds promise for addressing complex real-world applications...

Full description

Saved in:
Bibliographic Details
Published in:Applied soft computing Vol. 160; p. 111696
Main Authors: Deng, Lingyun, Liu, Sanyang
Format: Journal Article
Language:English
Published: Elsevier B.V 01.07.2024
Subjects:
ISSN:1568-4946, 1872-9681
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This work unveils design flaws within most metaheuristics, with a specific focus on issues associated with the arithmetic optimization algorithm (AOA). Despite being a simple metaheuristic optimizer inspired by mathematical operations, AOA holds promise for addressing complex real-world applications. However, a thorough analysis of its search mechanism reveals a heavy dependence on problem bounds for the quality of solutions obtained by AOA. Additionally, discrepancies between algorithm descriptions and implementations in AOA can mislead users and impede progress within the metaheuristic community. Experimental simulations conducted on various standard benchmarks including their shifted versions indicate a structural bias in AOA, leading to artificially high accuracy in fitting standard test functions but poor performance when applied to shifted benchmarks. Finally, we give a critical cause analysis and conclude this article by highlighting valuable research avenues in this field. •Defects of the arithmetic optimization algorithm (AOA) are revealed.•The search equation of AOA exhibits certain limitations.•AOA is structurally biased toward the origin of axes.
ISSN:1568-4946
1872-9681
DOI:10.1016/j.asoc.2024.111696