paper-with-me

Papers

Random Matrix Theory for Stochastic Gradient Descent

2024-12-29 · Chanju Park, Matteo Favoni, Biagio Lucini, Gert Aarts

Investigating the dynamics of learning in machine learning algorithms is of paramount importance for understanding how and why an approach may be successful. The tools of physics and statistics provide a robust setting for such investigations. Here we apply concepts from random matrix theory to describe stochastic weight matrix dynamics, using the framework of Dyson Brownian motion. We derive the linear scaling rule between the learning rate (step size) and the batch size, and identify universal and non-universal aspects of weight matrix dynamics. We test our findings in the (near-)solvable case of the Gaussian Restricted Boltzmann Machine and in a linear one-hidden-layer neural network.

📄 PDF Abstract BibTeX arXiv:2412.20496

Code (0)

등록된 구현이 없습니다.

Methods 이 논문이 사용한 방법론

Restricted Boltzmann Machine 설명 없음

Similar Papers 제목 키워드 기반

Analysing heavy-tail properties of Stochastic Gradient Descent by means of Stochastic Recurrence Equations

2024-03-20 · Ewa Damek, Sebastian Mentemeier

In recent works on the theory of machine learning, it has been observed that heavy tail properties of Stochastic Gradient Descent (SGD) can be studied in the probabilistic framework of stochastic recursions. In particula…

Learning Rates as a Function of Batch Size: A Random Matrix Theory Approach to Neural Network Training

2020-06-16 · Diego Granziol, Stefan Zohren, Stephen Roberts

We study the effect of mini-batching on the loss landscape of deep neural networks using spiked, field-dependent random matrix theory. We demonstrate that the magnitude of the extremal values of the batch Hessian are lar…

Second-order methods

Type-II Saddles and Probabilistic Stability of Stochastic Gradient Descent

2023-03-23 · Liu Ziyin, Botao Li, Tomer Galanti, Masahito Ueda

Characterizing and understanding the dynamics of stochastic gradient descent (SGD) around saddle points remains an open problem. We first show that saddle points in neural networks can be divided into two types, among wh…

Learning Theory

Global Convergence of Stochastic Gradient Descent for Some Non-convex Matrix Problems

2014-11-05 · Christopher De Sa, Kunle Olukotun, Christopher Ré

Stochastic gradient descent (SGD) on a low-rank factorization is commonly employed to speed up matrix problems including matrix completion, subspace tracking, and SDP relaxation. In this paper, we exhibit a step size sch…

Matrix Completion

A Markov Chain Theory Approach to Characterizing the Minimax Optimality of Stochastic Gradient Descent (for Least Squares)

2017-10-25 · Prateek Jain, Sham M. Kakade, Rahul Kidambi, Praneeth Netrapalli 외

This work provides a simplified proof of the statistical minimax optimality of (iterate averaged) stochastic gradient descent (SGD), for the special case of least squares. This result is obtained by analyzing SGD as a st…