The high-dimensional asymptotics of first order methods with random data

MaRDI QIDQ6385670zbMATH OpenFDO

Authors Michael Celentano, Chen Cheng, Andrea Montanari

Publication date 14 December 2021

Abstract: We study a class of deterministic flows in ${mathbbR}^{dimesk}$, parametrized by a random matrix $$ with i.i.d. centered subgaussian entries. We characterize the asymptotic behavior of these flows over bounded time horizons, in the high-dimensional limit in which $n,doinfty$ with $k$ fixed and converging aspect ratios $n/dodelta$. The asymptotic characterization we prove is in terms of a system of a nonlinear stochastic process in $k$ dimensions, whose parameters are determined by a fixed point condition. This type of characterization is known in physics as dynamical mean field theory. Rigorous results of this type have been obtained in the past for a few spin glass models. Our proof is based on time discretization and a reduction to certain iterative schemes known as approximate message passing (AMP) algorithms, as opposed to earlier work that was based on large deviations theory and stochastic processes theory. The new approach allows for a more elementary proof and implies that the high-dimensional behavior of the flow is universal with respect to the distribution of the entries of $$. As specific applications, we obtain high-dimensional characterizations of gradient flow in some classical models from statistics and machine learning, under a random design assumption.

This page was built for publication: The high-dimensional asymptotics of first order methods with random data

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6385670)