Automated design of search algorithms based on reinforcement learning

Yi, Wenjie; Qu, Rong

Abstract

Automated algorithm design has attracted increasing research attention recently in the evolutionary computation community. The main design decisions include selection heuristics and evolution operators in the search algorithms. Most existing studies, however, have focused on the automated design of evolution operators, neglecting selection heuristics for evolution and for replacement, not to mention considering all of the design decisions. This limited the scope of the algorithms under consideration. This study aims to systematically investigate automated design of search algorithms by exploring the impact of individual algorithmic components within a general search framework and the synergy among these multiple algorithmic components utilising a reinforcement learning technique. Comprehensive computational experiments are conducted on different benchmark instances of the capacitated vehicle routing problem with time windows to evaluate the effectiveness and generality of the proposed method. This study contributes to knowledge discovery in automated algorithm design using machine learning towards significantly enhanced generality of search algorithms.

• Journal Article Type: Article
• Acceptance Date: Aug 28, 2023
• Online Publication Date: Sep 1, 2023
• Publication Date: 2023-11
• Deposit Date: Sep 11, 2023
• Publicly Available Date: Sep 2, 2024
• Journal: Information Sciences
• Print ISSN: 0020-0255
• Publisher: Elsevier BV
• Peer Reviewed: Peer Reviewed
• Volume: 649
• Article Number: 119639
• DOI: https://doi.org/10.1016/j.ins.2023.119639
• Keywords: Artificial Intelligence; Information Systems and Management; Computer Science Applications; Theoretical Computer Science; Control and Systems Engineering; Software
• Public URL: https://nottingham-repository.worktribe.com/output/25083832
• Publisher URL: https://www.sciencedirect.com/science/article/abs/pii/S0020025523012240?via%3Dihub