Conservative set valued fields, automatic differentiation, stochastic gradient method and deep learning

TL;DR

This paper introduces conservative fields as a new framework for nonsmooth calculus, enabling variational formulas and convergence analysis for stochastic gradient methods in deep learning.

Contribution

It develops a calculus for conservative fields, providing a unified approach to nonsmooth automatic differentiation and convergence analysis in deep learning.

Findings

Established variational formulas for nonsmooth automatic differentiation

Proved convergence in values of nonsmooth stochastic gradient methods

Applied Whitney stratification to deep learning algorithms

Abstract

Modern problems in AI or in numerical analysis require nonsmooth approaches with a flexible calculus. We introduce generalized derivatives called conservative fields for which we develop a calculus and provide representation formulas. Functions having a conservative field are called path differentiable: convex, concave, Clarke regular and any semialgebraic Lipschitz continuous functions are path differentiable. Using Whitney stratification techniques for semialgebraic and definable sets, our model provides variational formulas for nonsmooth automatic differentiation oracles, as for instance the famous backpropagation algorithm in deep learning. Our differential model is applied to establish the convergence in values of nonsmooth stochastic gradient methods as they are implemented in practice.