A novel evolutionary status guided hyper-heuristic algorithm for continuous optimization

This paper proposes a novel evolutionary status guided hyper-heuristic algorithm named ES-HHA for continuous optimization. A representative hyper-heuristic algorithm consists of two components: the low-level component and the high-level component. In the low-level component, to balance the exploitat...

Full description

Saved in:
Bibliographic Details
Published in:Cluster computing Vol. 27; no. 9; pp. 12209 - 12238
Main Authors: Zhong, Rui, Yu, Jun
Format: Journal Article
Language:English
Published: New York Springer US 01.12.2024
Springer Nature B.V
Subjects:
Ablation
Computer Communication Networks
Computer Science
Design
Design optimization
Distance learning
Evolutionary algorithms
Evolutionary computation
Genetic algorithms
Heuristic
Heuristic methods
Mutation
Operating Systems
Operators (mathematics)
Optimization techniques
Performance evaluation
Processor Architectures
Statistical analysis
Task complexity
Hyper-heuristic algorithm
Evolutionary status
Low-level heuristics (LLHs)
Fitness distance correlation (FDC)
Population diversity (PD)
ISSN:1386-7857, 1573-7543
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper proposes a novel evolutionary status guided hyper-heuristic algorithm named ES-HHA for continuous optimization. A representative hyper-heuristic algorithm consists of two components: the low-level component and the high-level component. In the low-level component, to balance the exploitation and exploration during optimization, we design an exploitative operator pool and an explorative operator pool as low-level heuristics (LLHs), where the former is constructed using local search based operators, and the latter consists of various mutation operators from differential evolution (DE). In the high-level component, we design a probabilistic selection function based on the fitness distance correlation (FDC) and the population diversity (PD). Since these two metrics can reflect the complexity of the fitness landscape and the status of the evolutionary swarm, the integration of these two metrics is expected to determine the sequence of heuristics automatically and intelligently. To evaluate the performance of our proposal, we implement comprehensive numerical experiments on CEC2014, CEC2022, and eight engineering optimization tasks. A total of 14 famous optimization approaches are adopted as competitors. Furthermore, the ablation experiment is conducted to evaluate the high-level component independently, while the sensitivity experiment contributes to determining the optimal hyperparameter setting. The experimental results and statistical analysis show that ES-HHA is competitive, and the evolutionary status guided probabilistic selection function can determine the optimization intelligently.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1386-7857
1573-7543
DOI:10.1007/s10586-024-04593-2