Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Boursier, Etienne, Bowditch, Matthew, Englert, Matthias, Lazic, Ranko
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2505.22578
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910972897132544
author Boursier, Etienne
Bowditch, Matthew
Englert, Matthias
Lazic, Ranko
author_facet Boursier, Etienne
Bowditch, Matthew
Englert, Matthias
Lazic, Ranko
contents The optimization of neural networks under weight decay remains poorly understood from a theoretical standpoint. While weight decay is standard practice in modern training procedures, most theoretical analyses focus on unregularized settings. In this work, we investigate the loss landscape of the $\ell_2$-regularized training loss for two-layer ReLU networks. We show that the landscape becomes benign -- i.e., free of spurious local minima -- under large overparametrization, specifically when the network width $m$ satisfies $m \gtrsim \min(n^d, 2^n)$, where $n$ is the number of data points and $d$ the input dimension. More precisely in this regime, almost all constant activation regions contain a global minimum and no spurious local minima. We further show that this level of overparametrization is not only sufficient but also necessary via the example of orthogonal data. Finally, we demonstrate that such loss landscape results primarily hold relevance in the large initialization regime. In contrast, for small initializations -- corresponding to the feature learning regime -- optimization can still converge to spurious local minima, despite the global benignity of the landscape.
format Preprint
id arxiv_https___arxiv_org_abs_2505_22578
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Benignity of loss landscape with weight decay requires both large overparametrization and initialization
Boursier, Etienne
Bowditch, Matthew
Englert, Matthias
Lazic, Ranko
Machine Learning
The optimization of neural networks under weight decay remains poorly understood from a theoretical standpoint. While weight decay is standard practice in modern training procedures, most theoretical analyses focus on unregularized settings. In this work, we investigate the loss landscape of the $\ell_2$-regularized training loss for two-layer ReLU networks. We show that the landscape becomes benign -- i.e., free of spurious local minima -- under large overparametrization, specifically when the network width $m$ satisfies $m \gtrsim \min(n^d, 2^n)$, where $n$ is the number of data points and $d$ the input dimension. More precisely in this regime, almost all constant activation regions contain a global minimum and no spurious local minima. We further show that this level of overparametrization is not only sufficient but also necessary via the example of orthogonal data. Finally, we demonstrate that such loss landscape results primarily hold relevance in the large initialization regime. In contrast, for small initializations -- corresponding to the feature learning regime -- optimization can still converge to spurious local minima, despite the global benignity of the landscape.
title Benignity of loss landscape with weight decay requires both large overparametrization and initialization
topic Machine Learning
url https://arxiv.org/abs/2505.22578