Salvato in:
Dettagli Bibliografici
Autori principali: Galanti, Tomer, Siegel, Zachary S., Gupte, Aparna, Poggio, Tomaso
Natura: Preprint
Pubblicazione: 2022
Soggetti:
Accesso online:https://arxiv.org/abs/2206.05794
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909354253352960
author Galanti, Tomer
Siegel, Zachary S.
Gupte, Aparna
Poggio, Tomaso
author_facet Galanti, Tomer
Siegel, Zachary S.
Gupte, Aparna
Poggio, Tomaso
contents We investigate the inherent bias of Stochastic Gradient Descent (SGD) toward learning low-rank weight matrices during the training of deep neural networks. Our results demonstrate that training with mini-batch SGD and weight decay induces a bias toward rank minimization in the weight matrices. Specifically, we show both theoretically and empirically that this bias becomes more pronounced with smaller batch sizes, higher learning rates, or stronger weight decay. Additionally, we predict and empirically confirm that weight decay is essential for this bias to occur. Unlike previous literature, our analysis does not rely on assumptions about the data, convergence, or optimality of the weight matrices, making it applicable to a wide range of neural network architectures of any width or depth. Finally, we empirically explore the connection between this bias and generalization, finding that it has a marginal effect on the test performance.
format Preprint
id arxiv_https___arxiv_org_abs_2206_05794
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle SGD and Weight Decay Secretly Minimize the Rank of Your Neural Network
Galanti, Tomer
Siegel, Zachary S.
Gupte, Aparna
Poggio, Tomaso
Machine Learning
We investigate the inherent bias of Stochastic Gradient Descent (SGD) toward learning low-rank weight matrices during the training of deep neural networks. Our results demonstrate that training with mini-batch SGD and weight decay induces a bias toward rank minimization in the weight matrices. Specifically, we show both theoretically and empirically that this bias becomes more pronounced with smaller batch sizes, higher learning rates, or stronger weight decay. Additionally, we predict and empirically confirm that weight decay is essential for this bias to occur. Unlike previous literature, our analysis does not rely on assumptions about the data, convergence, or optimality of the weight matrices, making it applicable to a wide range of neural network architectures of any width or depth. Finally, we empirically explore the connection between this bias and generalization, finding that it has a marginal effect on the test performance.
title SGD and Weight Decay Secretly Minimize the Rank of Your Neural Network
topic Machine Learning
url https://arxiv.org/abs/2206.05794