Large-step GD near a flat-minima manifold of overparametrised least squares has a normal form that yields subcritical, critical, and supercritical convergence theorems, including for deep matrix factorisation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Dynamics of Gradient Descent with Large Step Size Near a Manifold of Flat Minima
Large-step GD near a flat-minima manifold of overparametrised least squares has a normal form that yields subcritical, critical, and supercritical convergence theorems, including for deep matrix factorisation.