On generalization bounds for deep networks based on loss surface implicit regularization

TL;DR

This paper investigates how the local geometry of the loss surface influences the generalization ability of deep neural networks, providing bounds that depend on spectral norms rather than parameter count.

Contribution

It introduces a new framework linking local loss surface geometry to implicit regularization, leading to tighter generalization bounds based on spectral norms.

Findings

SGD tends to stay near low-dimensional subspaces due to local geometry.

Generalization bounds depend on spectral norms, not parameter count.

Conditions for SGD stagnation imply improved generalization guarantees.

Abstract

The classical statistical learning theory implies that fitting too many parameters leads to overfitting and poor performance. That modern deep neural networks generalize well despite a large number of parameters contradicts this finding and constitutes a major unsolved problem towards explaining the success of deep learning. While previous work focuses on the implicit regularization induced by stochastic gradient descent (SGD), we study here how the local geometry of the energy landscape around local minima affects the statistical properties of SGD with Gaussian gradient noise. We argue that under reasonable assumptions, the local geometry forces SGD to stay close to a low dimensional subspace and that this induces another form of implicit regularization and results in tighter bounds on the generalization error for deep neural networks. To derive generalization error bounds for neural networks, we first introduce a notion of stagnation sets around the local minima and impose a local essential convexity property of the population risk. Under these conditions, lower bounds for SGD to remain in these stagnation sets are derived. If stagnation occurs, we derive a bound on the generalization error of deep neural networks involving the spectral norms of the weight matrices but not the number of network parameters. Technically, our proofs are based on controlling the change of parameter values in the SGD iterates and local uniform convergence of the empirical loss functions based on the entropy of suitable neighborhoods around local minima.