Provable Benefits of Unsupervised Pre-training and Transfer Learning via Single-Index Models

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Jones-McCormick, Taj, Jagannath, Aukosh, Sen, Subhabrata
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910999653646336
author Jones-McCormick, Taj
Jagannath, Aukosh
Sen, Subhabrata
author_facet Jones-McCormick, Taj
Jagannath, Aukosh
Sen, Subhabrata
contents Unsupervised pre-training and transfer learning are commonly used techniques to initialize training algorithms for neural networks, particularly in settings with limited labeled data. In this paper, we study the effects of unsupervised pre-training and transfer learning on the sample complexity of high-dimensional supervised learning. Specifically, we consider the problem of training a single-layer neural network via online stochastic gradient descent. We establish that pre-training and transfer learning (under concept shift) reduce sample complexity by polynomial factors (in the dimension) under very general assumptions. We also uncover some surprising settings where pre-training grants exponential improvement over random initialization in terms of sample complexity.
format Preprint
id arxiv_https___arxiv_org_abs_2502_16849
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Provable Benefits of Unsupervised Pre-training and Transfer Learning via Single-Index Models
Jones-McCormick, Taj
Jagannath, Aukosh
Sen, Subhabrata
Machine Learning
Unsupervised pre-training and transfer learning are commonly used techniques to initialize training algorithms for neural networks, particularly in settings with limited labeled data. In this paper, we study the effects of unsupervised pre-training and transfer learning on the sample complexity of high-dimensional supervised learning. Specifically, we consider the problem of training a single-layer neural network via online stochastic gradient descent. We establish that pre-training and transfer learning (under concept shift) reduce sample complexity by polynomial factors (in the dimension) under very general assumptions. We also uncover some surprising settings where pre-training grants exponential improvement over random initialization in terms of sample complexity.
title Provable Benefits of Unsupervised Pre-training and Transfer Learning via Single-Index Models
topic Machine Learning
url https://arxiv.org/abs/2502.16849