PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization

In this paper, we propose a novel stochastic gradient estimator---ProbAbilistic Gradient Estimator (PAGE)---for nonconvex optimization. PAGE is easy to implement as it is designed via a small adjustment to vanilla SGD: in each iteration, PAGE uses the vanilla minibatch SGD update with probability $p...

Full description

Saved in:
Bibliographic Details
Main Authors: LI, Zhize, BAO, Hongyan, ZHANG, Xiangliang, RICHTARIK, Peter
Format: text
Language:English
Published: Institutional Knowledge at Singapore Management University 2021
Subjects:
Online Access:https://ink.library.smu.edu.sg/sis_research/8683
https://ink.library.smu.edu.sg/context/sis_research/article/9686/viewcontent/ICML21_full_page.pdf
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Singapore Management University
Language: English
id sg-smu-ink.sis_research-9686
record_format dspace
spelling sg-smu-ink.sis_research-96862024-03-28T09:03:59Z PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization LI, Zhize BAO, Hongyan ZHANG, Xiangliang RICHTARIK, Peter In this paper, we propose a novel stochastic gradient estimator---ProbAbilistic Gradient Estimator (PAGE)---for nonconvex optimization. PAGE is easy to implement as it is designed via a small adjustment to vanilla SGD: in each iteration, PAGE uses the vanilla minibatch SGD update with probability $p_t$ or reuses the previous gradient with a small adjustment, at a much lower computational cost, with probability $1-p_t$. We give a simple formula for the optimal choice of $p_t$. Moreover, we prove the first tight lower bound $\Omega(n+\frac{\sqrt{n}}{\epsilon^2})$ for nonconvex finite-sum problems, which also leads to a tight lower bound $\Omega(b+\frac{\sqrt{b}}{\epsilon^2})$ for nonconvex online problems, where $b:= \min\{\frac{\sigma^2}{\epsilon^2}, n\}$. Then, we show that PAGE obtains the optimal convergence results $O(n+\frac{\sqrt{n}}{\epsilon^2})$ (finite-sum) and $O(b+\frac{\sqrt{b}}{\epsilon^2})$ (online) matching our lower bounds for both nonconvex finite-sum and online problems. Besides, we also show that for nonconvex functions satisfying the Polyak-\L ojasiewicz (PL) condition, PAGE can automatically switch to a faster linear convergence rate $O(\cdot\log \frac{1}{\epsilon})$. Finally, we conduct several deep learning experiments (e.g., LeNet, VGG, ResNet) on real datasets in PyTorch showing that PAGE not only converges much faster than SGD in training but also achieves the higher test accuracy, validating the optimal theoretical results and confirming the practical superiority of PAGE. 2021-07-01T07:00:00Z text application/pdf https://ink.library.smu.edu.sg/sis_research/8683 https://ink.library.smu.edu.sg/context/sis_research/article/9686/viewcontent/ICML21_full_page.pdf http://creativecommons.org/licenses/by-nc-nd/4.0/ Research Collection School Of Computing and Information Systems eng Institutional Knowledge at Singapore Management University Databases and Information Systems
institution Singapore Management University
building SMU Libraries
continent Asia
country Singapore
Singapore
content_provider SMU Libraries
collection InK@SMU
language English
topic Databases and Information Systems
spellingShingle Databases and Information Systems
LI, Zhize
BAO, Hongyan
ZHANG, Xiangliang
RICHTARIK, Peter
PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization
description In this paper, we propose a novel stochastic gradient estimator---ProbAbilistic Gradient Estimator (PAGE)---for nonconvex optimization. PAGE is easy to implement as it is designed via a small adjustment to vanilla SGD: in each iteration, PAGE uses the vanilla minibatch SGD update with probability $p_t$ or reuses the previous gradient with a small adjustment, at a much lower computational cost, with probability $1-p_t$. We give a simple formula for the optimal choice of $p_t$. Moreover, we prove the first tight lower bound $\Omega(n+\frac{\sqrt{n}}{\epsilon^2})$ for nonconvex finite-sum problems, which also leads to a tight lower bound $\Omega(b+\frac{\sqrt{b}}{\epsilon^2})$ for nonconvex online problems, where $b:= \min\{\frac{\sigma^2}{\epsilon^2}, n\}$. Then, we show that PAGE obtains the optimal convergence results $O(n+\frac{\sqrt{n}}{\epsilon^2})$ (finite-sum) and $O(b+\frac{\sqrt{b}}{\epsilon^2})$ (online) matching our lower bounds for both nonconvex finite-sum and online problems. Besides, we also show that for nonconvex functions satisfying the Polyak-\L ojasiewicz (PL) condition, PAGE can automatically switch to a faster linear convergence rate $O(\cdot\log \frac{1}{\epsilon})$. Finally, we conduct several deep learning experiments (e.g., LeNet, VGG, ResNet) on real datasets in PyTorch showing that PAGE not only converges much faster than SGD in training but also achieves the higher test accuracy, validating the optimal theoretical results and confirming the practical superiority of PAGE.
format text
author LI, Zhize
BAO, Hongyan
ZHANG, Xiangliang
RICHTARIK, Peter
author_facet LI, Zhize
BAO, Hongyan
ZHANG, Xiangliang
RICHTARIK, Peter
author_sort LI, Zhize
title PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization
title_short PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization
title_full PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization
title_fullStr PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization
title_full_unstemmed PAGE: A simple and optimal probabilistic gradient estimator for nonconvex optimization
title_sort page: a simple and optimal probabilistic gradient estimator for nonconvex optimization
publisher Institutional Knowledge at Singapore Management University
publishDate 2021
url https://ink.library.smu.edu.sg/sis_research/8683
https://ink.library.smu.edu.sg/context/sis_research/article/9686/viewcontent/ICML21_full_page.pdf
_version_ 1795302170978942976