A Mean-Field Optimal Control Formulation of Deep Learning

TL;DR

This paper formulates deep learning as a mean-field optimal control problem, providing a rigorous mathematical foundation and optimality conditions that connect population risk minimization with empirical learning.

Contribution

It introduces a systematic mean-field optimal control framework for deep learning, deriving optimality conditions and linking population and empirical risks.

Findings

Established Hamilton-Jacobi-Bellman and Pontryagin optimality conditions.

Derived quantitative relationships between population and empirical learning.

Provided a mathematical foundation for deep learning as an optimal control problem.

Abstract

Recent work linking deep neural networks and dynamical systems opened up new avenues to analyze deep learning. In particular, it is observed that new insights can be obtained by recasting deep learning as an optimal control problem on difference or differential equations. However, the mathematical aspects of such a formulation have not been systematically explored. This paper introduces the mathematical formulation of the population risk minimization problem in deep learning as a mean-field optimal control problem. Mirroring the development of classical optimal control, we state and prove optimality conditions of both the Hamilton-Jacobi-Bellman type and the Pontryagin type. These mean-field results reflect the probabilistic nature of the learning problem. In addition, by appealing to the mean-field Pontryagin's maximum principle, we establish some quantitative relationships between population and empirical learning problems. This serves to establish a mathematical foundation for investigating the algorithmic and theoretical connections between optimal control and deep learning.