High probability generalization bounds for uniformly stable algorithms with nearly optimal rate

TL;DR

This paper establishes nearly optimal high-probability generalization bounds for uniformly stable algorithms, including stochastic gradient descent and regularized ERM, improving upon previous bounds and resolving open problems.

Contribution

The authors prove a nearly tight high-probability generalization bound for uniformly stable algorithms, applicable to multi-pass SGD and convex ERM, with a novel proof technique.

Findings

Bound of O(γ log(n) log(n/δ) + √(log(1/δ)/n)) on estimation error

First high-probability bounds for multi-pass SGD in convex settings

Nearly optimal rate matching sampling error for γ = O(1/√n)

Abstract

Algorithmic stability is a classical approach to understanding and analysis of the generalization error of learning algorithms. A notable weakness of most stability-based generalization bounds is that they hold only in expectation. Generalization with high probability has been established in a landmark paper of Bousquet and Elisseeff (2002) albeit at the expense of an additional $\sqrt{n}$ factor in the bound. Specifically, their bound on the estimation error of any $\gamma$-uniformly stable learning algorithm on $n$ samples and range in $[0,1]$ is $O(\gamma \sqrt{n \log(1/\delta)} + \sqrt{\log(1/\delta)/n})$ with probability $\geq 1-\delta$. The $\sqrt{n}$ overhead makes the bound vacuous in the common settings where $\gamma \geq 1/\sqrt{n}$. A stronger bound was recently proved by the authors (Feldman and Vondrak, 2018) that reduces the overhead to at most $O(n^{1/4})$. Still, both of these results give optimal generalization bounds only when $\gamma = O(1/n)$. We prove a nearly tight bound of $O(\gamma \log(n)\log(n/\delta) + \sqrt{\log(1/\delta)/n})$ on the estimation error of any $\gamma$-uniformly stable algorithm. It implies that for algorithms that are uniformly stable with $\gamma = O(1/\sqrt{n})$, estimation error is essentially the same as the sampling error. Our result leads to the first high-probability generalization bounds for multi-pass stochastic gradient descent and regularized ERM for stochastic convex problems with nearly optimal rate --- resolving open problems in prior work. Our proof technique is new and we introduce several analysis tools that might find additional applications.