Supervised Quadratic Feature Analysis: Information Geometry for Dimensionality Reduction
Daniel Herrera-Esposito [email protected], Johannes Burge [email protected] · 2025
Read Supervised Quadratic Feature Analysis: Information Geometry for Dimensionality Reduction by Daniel Herrera-Esposito [email protected], Johannes Burge [email protected] (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-2502-00168.txt.
Description
Daniel Herrera-Esposito and Johannes Burge present Supervised Quadratic Feature Analysis for dimensionality reduction. Their 2025 arXiv paper seeks features that keep classes well separated. Labeled data are mapped into a smaller feature space. A common plan is to maximize gaps between class distributions. Information geometry offers a fresh way to measure those gaps.
Distributions become points on a curved statistical manifold in this view. The Fisher information metric sets distances along that manifold. Its geodesic distance is called the Fisher-Rao distance. The new method learns linear features that stretch that distance. Gaussian assumptions shape the class distributions in the model. Tests on real datasets check how well separation survives compression. The approach gives feature learning a geometric footing.
Catalog record
| CREATOR | Daniel Herrera-Esposito [email protected], Johannes Burge [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 100 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-00168