Improving clinical interpretability of linear neuroimaging models through feature whitening
Sara Petiton, Antoine Grigis, Raphaël Vock, Edouard Duchesnay · 2026
Read Improving clinical interpretability of linear neuroimaging models through feature whitening by Sara Petiton, Antoine Grigis, Raphaël Vock, Edouard Duchesnay (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-2604-20675.txt.
Description
Petiton, Grigis, Vock, and Duchesnay improve medical clarity of linear brain-scan models through feature whitening. Linked brain regions make weights reflect shared rather than own-region shares. They apply whitening to region groups with shared variance, plus a variant with tuned strength. They test region marks in two mental sorting tasks, telling bipolar illness or schizophrenia from healthy controls. Unlike PCA or ICA, which use whitening to reduce axes, their approach keeps the full input signal, suiting reading over choice. Whitening lifts weight clarity while keeping forecast scores.
Catalog record
| CREATOR | Sara Petiton, Antoine Grigis, Raphaël Vock, Edouard Duchesnay |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 19589 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.20675