Learning generative models across incomparable spaces
© 36th International Conference on Machine Learning, ICML 2019. All rights reserved. Generative Adversarial Networks have shown remarkable success in learning a distribution that faithfully recovers a reference distribution in its entirety. However, in some cases, we may want to only learn some aspe...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
2021
|
| Online Access: | https://hdl.handle.net/1721.1/132307 |
| _version_ | 1811077403709014016 |
|---|---|
| author | Bunne, C Alvarez-Melis, D Krause, A Jegelka, S |
| author_facet | Bunne, C Alvarez-Melis, D Krause, A Jegelka, S |
| author_sort | Bunne, C |
| collection | MIT |
| description | © 36th International Conference on Machine Learning, ICML 2019. All rights reserved. Generative Adversarial Networks have shown remarkable success in learning a distribution that faithfully recovers a reference distribution in its entirety. However, in some cases, we may want to only learn some aspects (e.g., cluster or manifold structure), while modifying others (e.g., style, orientation or dimension). In this work, we propose an approach to learn generative models across such incomparable spaces, and demonstrate how to steer the learned distribution towards target properties. A key component of our model is the Gromov-Wasserstein distance, a notion of discrepancy that compares distributions relationally rather than absolutely. While this framework subsumes current generative models in identically reproducing distributions, its inherent flexibility allows application to tasks in manifold learning, relational learning and cross-domain learning. |
| first_indexed | 2024-09-23T10:42:10Z |
| format | Article |
| id | mit-1721.1/132307 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T10:42:10Z |
| publishDate | 2021 |
| record_format | dspace |
| spelling | mit-1721.1/1323072021-09-21T03:48:10Z Learning generative models across incomparable spaces Bunne, C Alvarez-Melis, D Krause, A Jegelka, S © 36th International Conference on Machine Learning, ICML 2019. All rights reserved. Generative Adversarial Networks have shown remarkable success in learning a distribution that faithfully recovers a reference distribution in its entirety. However, in some cases, we may want to only learn some aspects (e.g., cluster or manifold structure), while modifying others (e.g., style, orientation or dimension). In this work, we propose an approach to learn generative models across such incomparable spaces, and demonstrate how to steer the learned distribution towards target properties. A key component of our model is the Gromov-Wasserstein distance, a notion of discrepancy that compares distributions relationally rather than absolutely. While this framework subsumes current generative models in identically reproducing distributions, its inherent flexibility allows application to tasks in manifold learning, relational learning and cross-domain learning. 2021-09-20T18:21:46Z 2021-09-20T18:21:46Z 2020-12-21T19:44:24Z Article http://purl.org/eprint/type/ConferencePaper https://hdl.handle.net/1721.1/132307 en http://proceedings.mlr.press/v97/bunne19a 36th International Conference on Machine Learning, ICML 2019 Creative Commons Attribution-Noncommercial-Share Alike http://creativecommons.org/licenses/by-nc-sa/4.0/ application/pdf arXiv |
| spellingShingle | Bunne, C Alvarez-Melis, D Krause, A Jegelka, S Learning generative models across incomparable spaces |
| title | Learning generative models across incomparable spaces |
| title_full | Learning generative models across incomparable spaces |
| title_fullStr | Learning generative models across incomparable spaces |
| title_full_unstemmed | Learning generative models across incomparable spaces |
| title_short | Learning generative models across incomparable spaces |
| title_sort | learning generative models across incomparable spaces |
| url | https://hdl.handle.net/1721.1/132307 |
| work_keys_str_mv | AT bunnec learninggenerativemodelsacrossincomparablespaces AT alvarezmelisd learninggenerativemodelsacrossincomparablespaces AT krausea learninggenerativemodelsacrossincomparablespaces AT jegelkas learninggenerativemodelsacrossincomparablespaces |