Abstract
Machine learning allows learning accurate but inadmissible heuristics for hard combinatorial puzzles like the 15-puzzle, the 24-puzzle, and Rubik's cube. In this paper, we investigate how to exploit these learned heuristics in the context of heuristic search with suboptimality guarantees. Specifically, we study how Focal Search (FS), a well-known bounded-suboptimal search algorithm can be modified to better exploit inadmissible learned heuristics. We propose to use Focal Discrepancy Search (FDS) in the context of learned heuristics, which uses a discrepancy function, instead of the learned heuristic, to sort the focal list. In our empirical evaluation, we evaluate FS and FDS using DeepCubeA, an effective learned heuristic for the 15-puzzle. We show that FDS substantially outperforms FS. This suggests that in some domains, when a highly accurate heuristics is available, one should always consider using discrepancies for better search.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: Proceedings of the International Symposium on Combinatorial Search
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
