Sequence to better sequence: Continuous revision of combinatorial structures
© 2017 by the author(s). We present a model that, after learning on observations of (sequence, outcome) pairs, can be efficiently used to revise a new sequence in order to improve its associated outcome. Our framework requires neither example improvements, nor additional evaluation of outcomes for p...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
2021
|
| Online Access: | https://hdl.handle.net/1721.1/137633 |
| _version_ | 1826194758688571392 |
|---|---|
| author | Jaakkola, Tommi Gifford, David Mueller, Jonas |
| author2 | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory |
| author_facet | Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory Jaakkola, Tommi Gifford, David Mueller, Jonas |
| author_sort | Jaakkola, Tommi |
| collection | MIT |
| description | © 2017 by the author(s). We present a model that, after learning on observations of (sequence, outcome) pairs, can be efficiently used to revise a new sequence in order to improve its associated outcome. Our framework requires neither example improvements, nor additional evaluation of outcomes for proposed revisions. To avoid combinatorial-search over sequence elements, we specify a generative model with continuous latent factors, which is learned via joint approximate inference using a recurrent variational autoencoder (VAE) and an outcome-predicting neural network module. Under this model, gradient methods can be used to efficiently optimize the continuous latent factors with respect to inferred outcomes. By appropriately constraining this optimization and using the VAE decoder to generate a revised sequence, we ensure the revision is fundamentally similar to the original sequence, is associated with better outcomes, and looks natural. These desiderata are proven to hold with high probability under our approach, which is empirically demonstrated for revising natural language sentences. |
| first_indexed | 2024-09-23T10:01:40Z |
| format | Article |
| id | mit-1721.1/137633 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T10:01:40Z |
| publishDate | 2021 |
| record_format | dspace |
| spelling | mit-1721.1/1376332023-04-07T19:45:52Z Sequence to better sequence: Continuous revision of combinatorial structures Jaakkola, Tommi Gifford, David Mueller, Jonas Massachusetts Institute of Technology. Computer Science and Artificial Intelligence Laboratory © 2017 by the author(s). We present a model that, after learning on observations of (sequence, outcome) pairs, can be efficiently used to revise a new sequence in order to improve its associated outcome. Our framework requires neither example improvements, nor additional evaluation of outcomes for proposed revisions. To avoid combinatorial-search over sequence elements, we specify a generative model with continuous latent factors, which is learned via joint approximate inference using a recurrent variational autoencoder (VAE) and an outcome-predicting neural network module. Under this model, gradient methods can be used to efficiently optimize the continuous latent factors with respect to inferred outcomes. By appropriately constraining this optimization and using the VAE decoder to generate a revised sequence, we ensure the revision is fundamentally similar to the original sequence, is associated with better outcomes, and looks natural. These desiderata are proven to hold with high probability under our approach, which is empirically demonstrated for revising natural language sentences. 2021-11-08T12:44:49Z 2021-11-08T12:44:49Z 2017 2019-05-29T14:20:31Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/137633 Jaakkola, Tommi, Gifford, David and Mueller, Jonas. 2017. "Sequence to better sequence: Continuous revision of combinatorial structures." en http://proceedings.mlr.press/v70/mueller17a.html Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf MIT web domain |
| spellingShingle | Jaakkola, Tommi Gifford, David Mueller, Jonas Sequence to better sequence: Continuous revision of combinatorial structures |
| title | Sequence to better sequence: Continuous revision of combinatorial structures |
| title_full | Sequence to better sequence: Continuous revision of combinatorial structures |
| title_fullStr | Sequence to better sequence: Continuous revision of combinatorial structures |
| title_full_unstemmed | Sequence to better sequence: Continuous revision of combinatorial structures |
| title_short | Sequence to better sequence: Continuous revision of combinatorial structures |
| title_sort | sequence to better sequence continuous revision of combinatorial structures |
| url | https://hdl.handle.net/1721.1/137633 |
| work_keys_str_mv | AT jaakkolatommi sequencetobettersequencecontinuousrevisionofcombinatorialstructures AT gifforddavid sequencetobettersequencecontinuousrevisionofcombinatorialstructures AT muellerjonas sequencetobettersequencecontinuousrevisionofcombinatorialstructures |