Learning with Deictic Representation
Most reinforcement learning methods operate on propositional representations of the world state. Such representations are often intractably large and generalize poorly. Using a deictic representation is believed to be a viable alternative: they promise generalization while allowing the use of e...
| Main Authors: | , , , |
|---|---|
| Language: | en_US |
| Published: |
2004
|
| Subjects: | |
| Online Access: | http://hdl.handle.net/1721.1/6685 |
| _version_ | 1811086009535823872 |
|---|---|
| author | Finney, Sarah Gardiol, Natalia H. Kaelbling, Leslie Pack Oates, Tim |
| author_facet | Finney, Sarah Gardiol, Natalia H. Kaelbling, Leslie Pack Oates, Tim |
| author_sort | Finney, Sarah |
| collection | MIT |
| description | Most reinforcement learning methods operate on propositional representations of the world state. Such representations are often intractably large and generalize poorly. Using a deictic representation is believed to be a viable alternative: they promise generalization while allowing the use of existing reinforcement-learning methods. Yet, there are few experiments on learning with deictic representations reported in the literature. In this paper we explore the effectiveness of two forms of deictic representation and a naive propositional representation in a simple blocks-world domain. We find, empirically, that the deictic representations actually worsen performance. We conclude with a discussion of possible causes of these results and strategies for more effective learning in domains with objects. |
| first_indexed | 2024-09-23T13:19:27Z |
| id | mit-1721.1/6685 |
| institution | Massachusetts Institute of Technology |
| language | en_US |
| last_indexed | 2024-09-23T13:19:27Z |
| publishDate | 2004 |
| record_format | dspace |
| spelling | mit-1721.1/66852019-04-11T02:52:50Z Learning with Deictic Representation Finney, Sarah Gardiol, Natalia H. Kaelbling, Leslie Pack Oates, Tim AI Reinforcement Learning Partial Observability Representations Most reinforcement learning methods operate on propositional representations of the world state. Such representations are often intractably large and generalize poorly. Using a deictic representation is believed to be a viable alternative: they promise generalization while allowing the use of existing reinforcement-learning methods. Yet, there are few experiments on learning with deictic representations reported in the literature. In this paper we explore the effectiveness of two forms of deictic representation and a naive propositional representation in a simple blocks-world domain. We find, empirically, that the deictic representations actually worsen performance. We conclude with a discussion of possible causes of these results and strategies for more effective learning in domains with objects. 2004-10-08T20:37:45Z 2004-10-08T20:37:45Z 2002-04-10 AIM-2002-006 http://hdl.handle.net/1721.1/6685 en_US AIM-2002-006 41 p. 5712208 bytes 1294450 bytes application/postscript application/pdf application/postscript application/pdf |
| spellingShingle | AI Reinforcement Learning Partial Observability Representations Finney, Sarah Gardiol, Natalia H. Kaelbling, Leslie Pack Oates, Tim Learning with Deictic Representation |
| title | Learning with Deictic Representation |
| title_full | Learning with Deictic Representation |
| title_fullStr | Learning with Deictic Representation |
| title_full_unstemmed | Learning with Deictic Representation |
| title_short | Learning with Deictic Representation |
| title_sort | learning with deictic representation |
| topic | AI Reinforcement Learning Partial Observability Representations |
| url | http://hdl.handle.net/1721.1/6685 |
| work_keys_str_mv | AT finneysarah learningwithdeicticrepresentation AT gardiolnataliah learningwithdeicticrepresentation AT kaelblinglesliepack learningwithdeicticrepresentation AT oatestim learningwithdeicticrepresentation |