Salvato in:
Dettagli Bibliografici
Autori principali: Wei, Yudong, Zhang, Liang, Li, Bingcong, He, Niao
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2510.01175
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912621188349952
author Wei, Yudong
Zhang, Liang
Li, Bingcong
He, Niao
author_facet Wei, Yudong
Zhang, Liang
Li, Bingcong
He, Niao
contents While normalization techniques are widely used in deep learning, their theoretical understanding remains relatively limited. In this work, we establish the benefits of (generalized) weight normalization (WN) applied to the overparameterized matrix sensing problem. We prove that WN with Riemannian optimization achieves linear convergence, yielding an exponential speedup over standard methods that do not use WN. Our analysis further demonstrates that both iteration and sample complexity improve polynomially as the level of overparameterization increases. To the best of our knowledge, this work provides the first characterization of how WN leverages overparameterization for faster convergence in matrix sensing.
format Preprint
id arxiv_https___arxiv_org_abs_2510_01175
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle On the Benefits of Weight Normalization for Overparameterized Matrix Sensing
Wei, Yudong
Zhang, Liang
Li, Bingcong
He, Niao
Machine Learning
Signal Processing
Optimization and Control
While normalization techniques are widely used in deep learning, their theoretical understanding remains relatively limited. In this work, we establish the benefits of (generalized) weight normalization (WN) applied to the overparameterized matrix sensing problem. We prove that WN with Riemannian optimization achieves linear convergence, yielding an exponential speedup over standard methods that do not use WN. Our analysis further demonstrates that both iteration and sample complexity improve polynomially as the level of overparameterization increases. To the best of our knowledge, this work provides the first characterization of how WN leverages overparameterization for faster convergence in matrix sensing.
title On the Benefits of Weight Normalization for Overparameterized Matrix Sensing
topic Machine Learning
Signal Processing
Optimization and Control
url https://arxiv.org/abs/2510.01175