ODE approximation for the Adam algorithm: General and overparametrized setting

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Dereich, Steffen, Jentzen, Arnulf, Kassing, Sebastian
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866914141325754368
author Dereich, Steffen
Jentzen, Arnulf
Kassing, Sebastian
author_facet Dereich, Steffen
Jentzen, Arnulf
Kassing, Sebastian
contents The Adam optimizer is currently presumably the most popular optimization method in deep learning. In this article we develop an ODE based method to study the Adam optimizer in a fast-slow scaling regime. For fixed momentum parameters and vanishing step-sizes, we show that the Adam algorithm is an asymptotic pseudo-trajectory of the flow of a particular vector field, which is referred to as the Adam vector field. Leveraging properties of asymptotic pseudo-trajectories, we establish convergence results for the Adam algorithm. In particular, in a very general setting we show that if the Adam algorithm converges, then the limit must be a zero of the Adam vector field, rather than a local minimizer or critical point of the objective function. In contrast, in the overparametrized empirical risk minimization setting, the Adam algorithm is able to locally find the set of minima. Specifically, we show that in a neighborhood of the global minima, the objective function serves as a Lyapunov function for the flow induced by the Adam vector field. As a consequence, if the Adam algorithm enters a neighborhood of the global minima infinitely often, it converges to the set of global minima.
format Preprint
id arxiv_https___arxiv_org_abs_2511_04622
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ODE approximation for the Adam algorithm: General and overparametrized setting
Dereich, Steffen
Jentzen, Arnulf
Kassing, Sebastian
Optimization and Control
Machine Learning
Probability
The Adam optimizer is currently presumably the most popular optimization method in deep learning. In this article we develop an ODE based method to study the Adam optimizer in a fast-slow scaling regime. For fixed momentum parameters and vanishing step-sizes, we show that the Adam algorithm is an asymptotic pseudo-trajectory of the flow of a particular vector field, which is referred to as the Adam vector field. Leveraging properties of asymptotic pseudo-trajectories, we establish convergence results for the Adam algorithm. In particular, in a very general setting we show that if the Adam algorithm converges, then the limit must be a zero of the Adam vector field, rather than a local minimizer or critical point of the objective function. In contrast, in the overparametrized empirical risk minimization setting, the Adam algorithm is able to locally find the set of minima. Specifically, we show that in a neighborhood of the global minima, the objective function serves as a Lyapunov function for the flow induced by the Adam vector field. As a consequence, if the Adam algorithm enters a neighborhood of the global minima infinitely often, it converges to the set of global minima.
title ODE approximation for the Adam algorithm: General and overparametrized setting
topic Optimization and Control
Machine Learning
Probability
url https://arxiv.org/abs/2511.04622