A Unified Matrix-Spectral Framework for Stability and Interpretability in Deep Learning
Ronald Katende · 2026
Read A Unified Matrix-Spectral Framework for Stability and Interpretability in Deep Learning by Ronald Katende (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-2602-01136.txt.
Description
Ronald Katende builds one matrix-spectrum frame for steadiness and read-ability in deep nets. Reading nets as data-tied chains of linear maps brings out spectrum numbers that steer touchiness to input tweaks, label noise, and train shifts. A global steadiness index pools Jacobians, weight slopes, tangent-kernel maps, and loss curves into one scale for forward touch, credit steadiness, and tune health. Tests on MNIST and CIFAR sets show mild spectrum tuning lifts credit steadiness a lot, even when broad spectrum sums shift little.
Catalog record
| CREATOR | Ronald Katende |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 30264 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602.01136