Critical Points of Degenerate Metrics on Algebraic Varieties: A Tale of Overparametrization
Giovanni Luca Marchetti *, Erin Connelly *, Paul Breiding *, Kathlén Kohn * · 2025
Read Critical Points of Degenerate Metrics on Algebraic Varieties: A Tale of Overparametrization by Giovanni Luca Marchetti *, Erin Connelly *, Paul Breiding *, Kathlén Kohn * (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-2512-21029.txt.
Description
Giovanni Luca Marchetti, Kathlen Kohn, and two co-authors bring algebraic geometry to the regime where models have more parameters than data. Fitting then means making a degenerate quadratic objective as small as possible over an algebraic variety, and the classical question of how many critical points the landscape has needs rework. Their main result connects the degenerate problem to a nondegenerate one through a projection, so existing counting tools can be carried across.
In the highly degenerate regime the ramification locus of that projection, where the projection folds, takes the central role. The authors supply tools for counting critical points over projective varieties and work through cases drawn from deep learning. The study extends the Euclidean distance degree field into the overparameterized ground where modern practice lives, giving the geometry of training landscapes a firmer footing.
Catalog record
| CREATOR | Giovanni Luca Marchetti *, Erin Connelly *, Paul Breiding *, Kathlén Kohn * |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 87 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-21029