Entropic Regularization in the Deep Linear Network

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chen, Alan, Kotwal, Tejas, Menon, Govind
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912752408199168
author Chen, Alan
Kotwal, Tejas
Menon, Govind
author_facet Chen, Alan
Kotwal, Tejas
Menon, Govind
contents We study regularization for the deep linear network (DLN) using the entropy formula introduced in arXiv:2509.09088. The equilibria and gradient flow of the free energy on the Riemannian manifold of end-to-end maps of the DLN are characterized for energies that depend symmetrically on the singular values of the end-to-end matrix. The only equilibria are minimizers and the set of minimizers is an orbit of the orthogonal group. In contrast with random matrix theory there is no singular value repulsion. The corresponding gradient flow reduces to a one-dimensional ordinary differential equation whose solution gives explicit relaxation rates toward the minimizers. We also study the concavity of the entropy in the chamber of singular values. The entropy is shown to be strictly concave in the Euclidean geometry on the chamber but not in the Riemannian geometry defined by the DLN metric.
format Preprint
id arxiv_https___arxiv_org_abs_2512_06137
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Entropic Regularization in the Deep Linear Network
Chen, Alan
Kotwal, Tejas
Menon, Govind
Neural and Evolutionary Computing
Dynamical Systems
Probability
37N40, 53C20, 15A18, 68T07
We study regularization for the deep linear network (DLN) using the entropy formula introduced in arXiv:2509.09088. The equilibria and gradient flow of the free energy on the Riemannian manifold of end-to-end maps of the DLN are characterized for energies that depend symmetrically on the singular values of the end-to-end matrix. The only equilibria are minimizers and the set of minimizers is an orbit of the orthogonal group. In contrast with random matrix theory there is no singular value repulsion. The corresponding gradient flow reduces to a one-dimensional ordinary differential equation whose solution gives explicit relaxation rates toward the minimizers. We also study the concavity of the entropy in the chamber of singular values. The entropy is shown to be strictly concave in the Euclidean geometry on the chamber but not in the Riemannian geometry defined by the DLN metric.
title Entropic Regularization in the Deep Linear Network
topic Neural and Evolutionary Computing
Dynamical Systems
Probability
37N40, 53C20, 15A18, 68T07
url https://arxiv.org/abs/2512.06137