Adversarial Dependence Minimization
Pierre-François De Plaen, Tinne Tuytelaars, Marc Proesmans, Luc Van Gool · 2025
Read Adversarial Dependence Minimization by Pierre-François De Plaen, Tinne Tuytelaars, Marc Proesmans, Luc Van Gool (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-2502-03227.txt.
Description
Pierre-François De Plaen, Tinne Tuytelaars, Marc Proesmans, and Luc Van Gool teach networks to keep features apart in a 2025 paper. Clean features should hold as little shared data as possible. Plain covariance cuts miss curved links between features. Their method, ADM, plays a game to cut all dependence.
Helper networks hunt for links between feature parts. The main encoder learns to erase whatever they find. At the best point, the team proves full independence of parts. Tests confirm the method converges in practice. Uses include a curved form of PCA and better image sorting. It also stops feature collapse in self-taught learning. The authors see gains for robust and compact models.
Catalog record
| CREATOR | Pierre-François De Plaen, Tinne Tuytelaars, Marc Proesmans, Luc Van Gool |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-03227