SuperCM: Improving Semi-Supervised Learning and Domain Adaptation through differentiable clustering Note: Visual Intelligence publications are financially supported by the Research Council of Norway, through its Center for Research-based Innovation funding scheme(grant no. 309439), and Consortium Pa
Durgesh Singh, Ahcène Boubekki, Robert Jenssen, Michael Kampffmeyer · 2025
Read SuperCM: Improving Semi-Supervised Learning and Domain Adaptation through differentiable clustering Note: Visual Intelligence publications are financially supported by the Research Council of Norway, through its Center for Research-based Innovation funding scheme(grant no. 309439), and Consortium Pa by Durgesh Singh, Ahcène Boubekki, Robert Jenssen, Michael Kampffmeyer (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-2507-13779.txt.
Description
Singh, Boubekki, Jenssen, and Kampffmeyer presented SuperCM in twenty twenty-five, a method improving learning when labels are scarce. Semi-supervised learning mixes a few labeled examples with many unlabeled ones. Domain adaptation applies the same idea when data shifts to a new setting.
Both tasks rest on a clustering idea: points grouped together should share a category. Past methods enforce this only indirectly during training. SuperCM instead builds clustering directly into the model as a module that can be trained end to end. Labeled data help set the cluster centers. Wide experiments show gains, mainly when supervision is very low. The method works alone or as an added regularizer that steadies existing approaches, the authors report.
Catalog record
| CREATOR | Durgesh Singh, Ahcène Boubekki, Robert Jenssen, Michael Kampffmeyer |
|---|---|
| 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/2507-13779