Overparameterization of deep ResNet: zero loss and mean-field analysis

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ding, Zhiyan, Chen, Shi, Li, Qin, Wright, Stephen
Format: Preprint
Published: 2021
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912261188091904
author Ding, Zhiyan
Chen, Shi
Li, Qin
Wright, Stephen
author_facet Ding, Zhiyan
Chen, Shi
Li, Qin
Wright, Stephen
contents Finding parameters in a deep neural network (NN) that fit training data is a nonconvex optimization problem, but a basic first-order optimization method (gradient descent) finds a global optimizer with perfect fit (zero-loss) in many practical situations. We examine this phenomenon for the case of Residual Neural Networks (ResNet) with smooth activation functions in a limiting regime in which both the number of layers (depth) and the number of weights in each layer (width) go to infinity. First, we use a mean-field-limit argument to prove that the gradient descent for parameter training becomes a gradient flow for a probability distribution that is characterized by a partial differential equation (PDE) in the large-NN limit. Next, we show that under certain assumptions, the solution to the PDE converges in the training time to a zero-loss solution. Together, these results suggest that the training of the ResNet gives a near-zero loss if the ResNet is large enough. We give estimates of the depth and width needed to reduce the loss below a given threshold, with high probability.
format Preprint
id arxiv_https___arxiv_org_abs_2105_14417
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle Overparameterization of deep ResNet: zero loss and mean-field analysis
Ding, Zhiyan
Chen, Shi
Li, Qin
Wright, Stephen
Machine Learning
Numerical Analysis
Finding parameters in a deep neural network (NN) that fit training data is a nonconvex optimization problem, but a basic first-order optimization method (gradient descent) finds a global optimizer with perfect fit (zero-loss) in many practical situations. We examine this phenomenon for the case of Residual Neural Networks (ResNet) with smooth activation functions in a limiting regime in which both the number of layers (depth) and the number of weights in each layer (width) go to infinity. First, we use a mean-field-limit argument to prove that the gradient descent for parameter training becomes a gradient flow for a probability distribution that is characterized by a partial differential equation (PDE) in the large-NN limit. Next, we show that under certain assumptions, the solution to the PDE converges in the training time to a zero-loss solution. Together, these results suggest that the training of the ResNet gives a near-zero loss if the ResNet is large enough. We give estimates of the depth and width needed to reduce the loss below a given threshold, with high probability.
title Overparameterization of deep ResNet: zero loss and mean-field analysis
topic Machine Learning
Numerical Analysis
url https://arxiv.org/abs/2105.14417