Stochastic Modified Flows, Mean-Field Limits and Dynamics of Stochastic Gradient Descent

02/14/2023
by   Benjamin Gess, et al.
0

We propose new limiting dynamics for stochastic gradient descent in the small learning rate regime called stochastic modified flows. These SDEs are driven by a cylindrical Brownian motion and improve the so-called stochastic modified equations by having regular diffusion coefficients and by matching the multi-point statistics. As a second contribution, we introduce distribution dependent stochastic modified flows which we prove to describe the fluctuating limiting dynamics of stochastic gradient descent in the small learning rate - infinite width scaling regime.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
07/12/2022

Conservative SPDEs as fluctuating mean field limits of stochastic gradient descent

The convergence of stochastic interacting particle systems in the mean-f...
research
12/18/2020

Convergence dynamics of Generative Adversarial Networks: the dual metric flows

Fitting neural networks often resorts to stochastic (or similar) gradien...
research
02/22/2022

A duality connecting neural network and cosmological dynamics

We demonstrate that the dynamics of neural networks trained with gradien...
research
06/19/2021

Learning and Generalization in Overparameterized Normalizing Flows

In supervised learning, it is known that overparameterized neural networ...
research
02/14/2020

Stochasticity of Deterministic Gradient Descent: Large Learning Rate for Multiscale Objective Function

This article suggests that deterministic Gradient Descent, which does no...
research
06/04/2021

Fluctuation-dissipation Type Theorem in Stochastic Linear Learning

The fluctuation-dissipation theorem (FDT) is a simple yet powerful conse...
research
06/25/2020

Automatic Tuning of Stochastic Gradient Descent with Bayesian Optimisation

Many machine learning models require a training procedure based on runni...

Please sign up or login with your details

Forgot password? Click here to reset