Gradient-Free Generation for Hard-Constrained Systems
Chaoran Cheng, Boran Han, Danielle C. Maddix, Abdul Fatir Ansari, Andrew Stuart, Michael W. Mahoney, Yuyang Wang · 2024
Read Gradient-Free Generation for Hard-Constrained Systems by Chaoran Cheng, Boran Han, Danielle C. Maddix, Abdul Fatir Ansari, Andrew Stuart, Michael W. Mahoney, Yuyang Wang (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-01786.txt.
Description
Chaoran Cheng and co-authors tackle a design problem in this 2024 paper: how to make AI generators obey hard rules. In science and engineering work, outputs often have to meet strict physical laws, and methods that lean on gradient data can be slow or costly there. The team offers a way to adapt a trained flow-matching model so its outputs meet the rules exactly, with no new training.
Their method, called ECI sampling, cycles through three steps at each stage of sampling: pushing the sample outward, correcting it, and blending it back. Tests on several systems governed by partial differential equations showed the outputs held to the rules strictly. The method also beat rival approaches on these tasks and did well on prediction tasks without extra tuning.
Catalog record
| CREATOR | Chaoran Cheng, Boran Han, Danielle C. Maddix, Abdul Fatir Ansari, Andrew Stuart, Michael W. Mahoney, Yuyang Wang |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 86 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-01786