Understanding the Implicit Regularization of Gradient Descent in Over-parameterized Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ma, Jianhao, Liang, Geyu, Fattahi, Salar
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911308977274880
author Ma, Jianhao
Liang, Geyu
Fattahi, Salar
author_facet Ma, Jianhao
Liang, Geyu
Fattahi, Salar
contents Implicit regularization refers to the tendency of local search algorithms to converge to low-dimensional solutions, even when such structures are not explicitly enforced. Despite its ubiquity, the mechanism underlying this behavior remains poorly understood, particularly in over-parameterized settings. We analyze gradient descent dynamics and identify three conditions under which it converges to second-order stationary points within an implicit low-dimensional region: (i) suitable initialization, (ii) efficient escape from saddle points, and (iii) sustained proximity to the region. We show that these can be achieved through infinitesimal perturbations and a small deviation rate. Building on this, we introduce Infinitesimally Perturbed Gradient Descent (IPGD), which satisfies these conditions under mild assumptions. We provide theoretical guarantees for IPGD in over-parameterized matrix sensing and empirical evidence of its broader applicability.
format Preprint
id arxiv_https___arxiv_org_abs_2505_17304
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Understanding the Implicit Regularization of Gradient Descent in Over-parameterized Models
Ma, Jianhao
Liang, Geyu
Fattahi, Salar
Machine Learning
Optimization and Control
Implicit regularization refers to the tendency of local search algorithms to converge to low-dimensional solutions, even when such structures are not explicitly enforced. Despite its ubiquity, the mechanism underlying this behavior remains poorly understood, particularly in over-parameterized settings. We analyze gradient descent dynamics and identify three conditions under which it converges to second-order stationary points within an implicit low-dimensional region: (i) suitable initialization, (ii) efficient escape from saddle points, and (iii) sustained proximity to the region. We show that these can be achieved through infinitesimal perturbations and a small deviation rate. Building on this, we introduce Infinitesimally Perturbed Gradient Descent (IPGD), which satisfies these conditions under mild assumptions. We provide theoretical guarantees for IPGD in over-parameterized matrix sensing and empirical evidence of its broader applicability.
title Understanding the Implicit Regularization of Gradient Descent in Over-parameterized Models
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2505.17304