Phase Transitions in Unsupervised Feature Selection
Jonathan Fiorentino *, † \dagger, Michele Monti *, Dimitrios Miltiadis-Vrachnos, Vittorio Del Tatto, Alessandro Laio, Gian Gaetano Tartaglia † \dagger · 2026
Read Phase Transitions in Unsupervised Feature Selection by Jonathan Fiorentino *, † \dagger, Michele Monti *, Dimitrios Miltiadis-Vrachnos, Vittorio Del Tatto, Alessandro Laio, Gian Gaetano Tartaglia † \dagger (2026) 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-2602-00660.txt.
Description
Jonathan Fiorentino and colleagues analyse unsupervised feature selection through the Differentiable Information Imbalance. The test case is proteins. Treating kept features as coordinates of a physics model, they show it passes through a phase transition as more features are kept. For physico-chemical descriptors the change runs from a glass-like phase with few features to a liquid-like phase. It is marked by two-peaked order-parameter patterns and Binder cumulant minima. The critical count matches where classification stops improving. This gives a clear rule for feature sets.
Catalog record
| CREATOR | Jonathan Fiorentino *, † \dagger, Michele Monti *, Dimitrios Miltiadis-Vrachnos, Vittorio Del Tatto, Alessandro Laio, Gian Gaetano Tartaglia † \dagger |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19129 |
| LICENSE | http://creativecommons.org/publicdomain/zero/1.0/ |
| RIGHTS | CC0 |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.00660