Learning to Match Unpaired Data with Minimum Entropy Coupling
Mustapha Bounoua, Giulio Franzese, Pietro Michiardi · 2025
Read Learning to Match Unpaired Data with Minimum Entropy Coupling by Mustapha Bounoua, Giulio Franzese, Pietro Michiardi (2025) 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-2503-08501.txt.
Description
Mustapha Bounoua, Giulio Franzese, and Pietro Michiardi gave a method to match unpaired data across modalities in two thousand twenty-five. Data in different forms, like images and text or two cell assays, often come without pair labels. Minimum entropy coupling seeks a joint distribution that matches the known margins while staying as simple as possible.
Past solvers handled only finite, discrete data. Their method uses diffusion generative models in a cooperative scheme to approximate and minimize joint entropy for continuous data, under relaxed margin rules. It aligned single-cell multi-omics data without supervision and translated images without pairs. On both tasks it beat specialized methods, which the authors cite as proof of its generality.
Catalog record
| CREATOR | Mustapha Bounoua, Giulio Franzese, Pietro Michiardi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-08501