Learning to rank for synthesizing planning heuristics
Author(s)
Garrett, Caelan Reed; Kaelbling, Leslie P; Lozano-Perez, Tomas
DownloadKaelbling_Learning to rank.pdf (301.8Kb)
OPEN_ACCESS_POLICY
Open Access Policy
Creative Commons Attribution-Noncommercial-Share Alike
Terms of use
Metadata
Show full item recordAbstract
We investigate learning heuristics for domainspecific planning. Prior work framed learning a heuristic as an ordinary regression problem. However,
in a greedy best-first search, the ordering of states induced by a heuristic is more indicative of the resulting planner’s performance than mean squared error. Thus, we instead frame learning a heuristic as a learning to rank problem which we solve using a RankSVM formulation. Additionally, we introduce new methods for computing features that capture temporal interactions in an approximate plan. Our experiments on recent International Planning Competition problems show that the RankSVM learned heuristics outperform both the original heuristics and heuristics learned through ordinary regression.
Date issued
2016-07Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer ScienceJournal
Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence (IJCAI-16)
Publisher
AAAI Press
Citation
Garrett, Caelan Reed, Leslie Pack Kaelbling, Tomás Lozano-Pérez. "Learning to Rank for Synthesizing Planning Heuristics." Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence (IJCAI-16), 9-15 July, 2016, New York, New York, AAAI Press.
Version: Author's final manuscript