Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Arnaboldi, Luca, Krzakala, Florent, Loureiro, Bruno, Stephan, Ludovic
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2305.18502
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866916143060484096
author Arnaboldi, Luca
Krzakala, Florent
Loureiro, Bruno
Stephan, Ludovic
author_facet Arnaboldi, Luca
Krzakala, Florent
Loureiro, Bruno
Stephan, Ludovic
contents This study explores the sample complexity for two-layer neural networks to learn a generalized linear target function under Stochastic Gradient Descent (SGD), focusing on the challenging regime where many flat directions are present at initialization. It is well-established that in this scenario $n=O(d \log d)$ samples are typically needed. However, we provide precise results concerning the pre-factors in high-dimensional contexts and for varying widths. Notably, our findings suggest that overparameterization can only enhance convergence by a constant factor within this problem class. These insights are grounded in the reduction of SGD dynamics to a stochastic process in lower dimensions, where escaping mediocrity equates to calculating an exit time. Yet, we demonstrate that a deterministic approximation of this process adequately represents the escape time, implying that the role of stochasticity may be minimal in this scenario.
format Preprint
id arxiv_https___arxiv_org_abs_2305_18502
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Escaping mediocrity: how two-layer networks learn hard generalized linear models with SGD
Arnaboldi, Luca
Krzakala, Florent
Loureiro, Bruno
Stephan, Ludovic
Machine Learning
This study explores the sample complexity for two-layer neural networks to learn a generalized linear target function under Stochastic Gradient Descent (SGD), focusing on the challenging regime where many flat directions are present at initialization. It is well-established that in this scenario $n=O(d \log d)$ samples are typically needed. However, we provide precise results concerning the pre-factors in high-dimensional contexts and for varying widths. Notably, our findings suggest that overparameterization can only enhance convergence by a constant factor within this problem class. These insights are grounded in the reduction of SGD dynamics to a stochastic process in lower dimensions, where escaping mediocrity equates to calculating an exit time. Yet, we demonstrate that a deterministic approximation of this process adequately represents the escape time, implying that the role of stochasticity may be minimal in this scenario.
title Escaping mediocrity: how two-layer networks learn hard generalized linear models with SGD
topic Machine Learning
url https://arxiv.org/abs/2305.18502