Breadth-First Search vs. Restarting Random Walks for Escaping Uninformed Heuristic Regions
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (43) · 37143-37151
DOI: 10.1609/aaai.v40i43.41044 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Daniel Platnick und 4 weiteren Autor:innen.
Abstract (Originalsprache)
Greedy search methods such as Greedy Best-First Search (GBFS) and Enforced Hill-Climbing (EHC) often struggle when faced with Uninformed Heuristic Regions (UHRs) like heuristic local minima or plateaus. In this work, we theoretically and empirically compare two popular methods for escaping UHRs: breadth-first search (BrFS) and restarting random walks (RRWs). First, we derive the expected runtime of escaping a UHR using BrFS and RRWs, based on properties of the UHR and the random walk procedure, and then use these results to identify when RRWs will be faster in expectation than BrFS. Next, we evaluate these methods for escaping UHRs by comparing standard EHC, which uses BrFS to escape UHRs, to variants of EHC called EHC-RRW, which use RRWs for that purpose. EHC-RRW is shown to have strong expected runtime guarantees in cases where EHC has previously been shown to be effective. Finally, we run experiments with these approaches on PDDL planning benchmarks to better understand their relative effectiveness for escaping UHRs.
Zitieren
APA
Platnick, D., Tomasz, D., Earl, E., Khanzadeh, S., & Valenzano, R. (2026). Breadth-First Search vs. Restarting Random Walks for Escaping Uninformed Heuristic Regions. Proceedings of the AAAI Conference on Artificial Intelligence, 40(43), 37143-37151. https://doi.org/10.1609/aaai.v40i43.41044
Vancouver
Platnick D, Tomasz D, Earl E, Khanzadeh S, Valenzano R. Breadth-First Search vs. Restarting Random Walks for Escaping Uninformed Heuristic Regions. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(43):37143-37151. doi:10.1609/aaai.v40i43.41044
BibTeX
@article{Platnick2026,
author = {Platnick, Daniel and Tomasz, Dawson and Earl, Eamon and Khanzadeh, Sourena and Valenzano, Richard},
title = {Breadth-First Search vs. Restarting Random Walks for Escaping Uninformed Heuristic Regions},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {43},
pages = {37143--37151},
doi = {10.1609/aaai.v40i43.41044}
}
EUVIMED