Modeling expectation violation in intuitive physics with coarse probabilistic object representations
© 2019 Neural information processing systems foundation. All rights reserved. From infancy, humans have expectations about how objects will move and interact. Even young children expect objects not to move through one another, teleport, or disappear. They are surprised by mismatches between physical...
| Main Authors: | , , , , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
2022
|
| Online Access: | https://hdl.handle.net/1721.1/138344.2 |
| _version_ | 1811093968335667200 |
|---|---|
| author | Smith, KA Mei, L Yao, S Wu, J Spelke, E Tenenbaum, JB Ullman, TD |
| author2 | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences |
| author_facet | Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences Smith, KA Mei, L Yao, S Wu, J Spelke, E Tenenbaum, JB Ullman, TD |
| author_sort | Smith, KA |
| collection | MIT |
| description | © 2019 Neural information processing systems foundation. All rights reserved. From infancy, humans have expectations about how objects will move and interact. Even young children expect objects not to move through one another, teleport, or disappear. They are surprised by mismatches between physical expectations and perceptual observations, even in unfamiliar scenes with completely novel objects. A model that exhibits human-like understanding of physics should be similarly surprised, and adjust its beliefs accordingly. We propose ADEPT, a model that uses a coarse (approximate geometry) object-centric representation for dynamic 3D scene understanding. Inference integrates deep recognition networks, extended probabilistic physical simulation, and particle filtering for forming predictions and expectations across occlusion. We also present a new test set for measuring violations of physical expectations, using a range of scenarios derived from developmental psychology. We systematically compare ADEPT, baseline models, and human expectations on this test set. ADEPT outperforms standard network architectures in discriminating physically implausible scenes, and often performs this discrimination at the same level as people. |
| first_indexed | 2024-09-23T15:53:30Z |
| format | Article |
| id | mit-1721.1/138344.2 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T15:53:30Z |
| publishDate | 2022 |
| record_format | dspace |
| spelling | mit-1721.1/138344.22024-06-03T17:26:09Z Modeling expectation violation in intuitive physics with coarse probabilistic object representations Smith, KA Mei, L Yao, S Wu, J Spelke, E Tenenbaum, JB Ullman, TD Massachusetts Institute of Technology. Department of Brain and Cognitive Sciences Center for Brains, Minds, and Machines Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory © 2019 Neural information processing systems foundation. All rights reserved. From infancy, humans have expectations about how objects will move and interact. Even young children expect objects not to move through one another, teleport, or disappear. They are surprised by mismatches between physical expectations and perceptual observations, even in unfamiliar scenes with completely novel objects. A model that exhibits human-like understanding of physics should be similarly surprised, and adjust its beliefs accordingly. We propose ADEPT, a model that uses a coarse (approximate geometry) object-centric representation for dynamic 3D scene understanding. Inference integrates deep recognition networks, extended probabilistic physical simulation, and particle filtering for forming predictions and expectations across occlusion. We also present a new test set for measuring violations of physical expectations, using a range of scenarios derived from developmental psychology. We systematically compare ADEPT, baseline models, and human expectations on this test set. ADEPT outperforms standard network architectures in discriminating physically implausible scenes, and often performs this discrimination at the same level as people. 2022-02-08T19:30:57Z 2021-12-07T14:13:15Z 2022-02-08T19:30:57Z 2019-01 2018-12 2021-12-07T14:10:39Z Article http://purl.org/eprint/type/ConferencePaper 9781510884472 https://hdl.handle.net/1721.1/138344.2 Smith, KA, Mei, L, Yao, S, Wu, J, Spelke, E et al. 2019. "Modeling expectation violation in intuitive physics with coarse probabilistic object representations." Advances in Neural Information Processing Systems, 32. en https://papers.nips.cc/paper/2019/hash/e88f243bf341ded9b4ced444795c3f17-Abstract.html Advances in Neural Information Processing Systems Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/octet-stream Neural Information Processing Systems (NIPS) |
| spellingShingle | Smith, KA Mei, L Yao, S Wu, J Spelke, E Tenenbaum, JB Ullman, TD Modeling expectation violation in intuitive physics with coarse probabilistic object representations |
| title | Modeling expectation violation in intuitive physics with coarse probabilistic object representations |
| title_full | Modeling expectation violation in intuitive physics with coarse probabilistic object representations |
| title_fullStr | Modeling expectation violation in intuitive physics with coarse probabilistic object representations |
| title_full_unstemmed | Modeling expectation violation in intuitive physics with coarse probabilistic object representations |
| title_short | Modeling expectation violation in intuitive physics with coarse probabilistic object representations |
| title_sort | modeling expectation violation in intuitive physics with coarse probabilistic object representations |
| url | https://hdl.handle.net/1721.1/138344.2 |
| work_keys_str_mv | AT smithka modelingexpectationviolationinintuitivephysicswithcoarseprobabilisticobjectrepresentations AT meil modelingexpectationviolationinintuitivephysicswithcoarseprobabilisticobjectrepresentations AT yaos modelingexpectationviolationinintuitivephysicswithcoarseprobabilisticobjectrepresentations AT wuj modelingexpectationviolationinintuitivephysicswithcoarseprobabilisticobjectrepresentations AT spelkee modelingexpectationviolationinintuitivephysicswithcoarseprobabilisticobjectrepresentations AT tenenbaumjb modelingexpectationviolationinintuitivephysicswithcoarseprobabilisticobjectrepresentations AT ullmantd modelingexpectationviolationinintuitivephysicswithcoarseprobabilisticobjectrepresentations |