Explaining generalization in deep learning: progress and fundamental limits

TL;DR

This paper investigates why deep networks generalize well despite overparameterization, analyzing uniform convergence limits, and proposing an empirical method using unlabeled data to estimate generalization.

Contribution

It provides empirical insights into implicit capacity control, derives improved data-dependent bounds, and introduces a novel empirical technique for estimating generalization without relying solely on uniform convergence.

Findings

Uniform convergence bounds can be vacuous in overparameterized settings.

Training via stochastic gradient descent implicitly controls network capacity.

An empirical method using unlabeled data accurately estimates generalization.

Abstract

This dissertation studies a fundamental open challenge in deep learning theory: why do deep networks generalize well even while being overparameterized, unregularized and fitting the training data to zero error? In the first part of the thesis, we will empirically study how training deep networks via stochastic gradient descent implicitly controls the networks' capacity. Subsequently, to show how this leads to better generalization, we will derive {\em data-dependent} {\em uniform-convergence-based} generalization bounds with improved dependencies on the parameter count. Uniform convergence has in fact been the most widely used tool in deep learning literature, thanks to its simplicity and generality. Given its popularity, in this thesis, we will also take a step back to identify the fundamental limits of uniform convergence as a tool to explain generalization. In particular, we will show that in some example overparameterized settings, {\em any} uniform convergence bound will provide only a vacuous generalization bound. With this realization in mind, in the last part of the thesis, we will change course and introduce an {\em empirical} technique to estimate generalization using unlabeled data. Our technique does not rely on any notion of uniform-convergece-based complexity and is remarkably precise. We will theoretically show why our technique enjoys such precision. We will conclude by discussing how future work could explore novel ways to incorporate distributional assumptions in generalization bounds (such as in the form of unlabeled data) and explore other tools to derive bounds, perhaps by modifying uniform convergence or by developing completely new tools altogether.