Introduces WSFN, a Newton-type method on Wasserstein space that escapes saddle points in polynomial time and achieves linear convergence to global minimizers under benign landscape assumptions.
Steffen Dereich, Arnulf Jentzen, and Sebastian Kassing
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
A λ-convex variational surrogate for shallow NN training yields global well-posedness, almost C³ regularity, and an explicit linear-system solution with 1/α generalization and O(1/N) finite-width rates.
citing papers explorer
-
From Saddle Points Toward Global Minima: A Newton-Type Method on Wasserstein Space
Introduces WSFN, a Newton-type method on Wasserstein space that escapes saddle points in polynomial time and achieves linear convergence to global minimizers under benign landscape assumptions.
-
Born Discrete, Made Smooth: Variational Formulation of Shallow Neural Networks
A λ-convex variational surrogate for shallow NN training yields global well-posedness, almost C³ regularity, and an explicit linear-system solution with 1/α generalization and O(1/N) finite-width rates.