Learning to rank for synthesizing planning heuristics
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...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | en_US |
| Published: |
AAAI Press
2018
|
| Online Access: | http://hdl.handle.net/1721.1/115313 https://orcid.org/0000-0002-6474-1276 https://orcid.org/0000-0001-6054-7145 https://orcid.org/0000-0002-8657-2450 |
| _version_ | 1826205052554969088 |
|---|---|
| author | Garrett, Caelan Reed Kaelbling, Leslie P Lozano-Perez, Tomas |
| author2 | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science |
| author_facet | Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Garrett, Caelan Reed Kaelbling, Leslie P Lozano-Perez, Tomas |
| author_sort | Garrett, Caelan Reed |
| collection | MIT |
| description | 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. |
| first_indexed | 2024-09-23T13:06:00Z |
| format | Article |
| id | mit-1721.1/115313 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T13:06:00Z |
| publishDate | 2018 |
| publisher | AAAI Press |
| record_format | dspace |
| spelling | mit-1721.1/1153132022-10-01T13:03:24Z Learning to rank for synthesizing planning heuristics Garrett, Caelan Reed Kaelbling, Leslie P Lozano-Perez, Tomas Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science Garrett, Caelan Reed Kaelbling, Leslie P Lozano-Perez, Tomas 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. National Science Foundation (U.S.) (Grant 1420927) National Science Foundation (U.S.) (Grant 1420927) National Science Foundation (U.S.) (Grant 1523767) United States. Office of Naval Research (Grant N00014-14-1-0486) United States. Army Research Office (Grant W911NF1410433) 2018-05-11T14:19:54Z 2018-05-11T14:19:54Z 2016-07 Article http://purl.org/eprint/type/ConferencePaper http://hdl.handle.net/1721.1/115313 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. https://orcid.org/0000-0002-6474-1276 https://orcid.org/0000-0001-6054-7145 https://orcid.org/0000-0002-8657-2450 en_US https://www.ijcai.org/proceedings/2016 Proceedings of the Twenty-Fifth International Joint Conference on Artificial Intelligence (IJCAI-16) Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by/4.0/ application/pdf AAAI Press MIT Web Domain |
| spellingShingle | Garrett, Caelan Reed Kaelbling, Leslie P Lozano-Perez, Tomas Learning to rank for synthesizing planning heuristics |
| title | Learning to rank for synthesizing planning heuristics |
| title_full | Learning to rank for synthesizing planning heuristics |
| title_fullStr | Learning to rank for synthesizing planning heuristics |
| title_full_unstemmed | Learning to rank for synthesizing planning heuristics |
| title_short | Learning to rank for synthesizing planning heuristics |
| title_sort | learning to rank for synthesizing planning heuristics |
| url | http://hdl.handle.net/1721.1/115313 https://orcid.org/0000-0002-6474-1276 https://orcid.org/0000-0001-6054-7145 https://orcid.org/0000-0002-8657-2450 |
| work_keys_str_mv | AT garrettcaelanreed learningtorankforsynthesizingplanningheuristics AT kaelblinglesliep learningtorankforsynthesizingplanningheuristics AT lozanopereztomas learningtorankforsynthesizingplanningheuristics |