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

Vitaly Feldman, Jan Vondrak
Proceedings of the Thirty-Second Conference on Learning Theory, PMLR 99:1270-1279, 2019.

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 (2001) albeit at the expense of an additional √n factor in the bound. Specifically, their bound on the estimation error of any γ-uniformly stable learning algorithm on n samples and range in [0,1] is O(γ√nlog(1/δ)+√log(1/δ)/n) with probability ≥1−δ. The √n overhead makes the bound vacuous in the common settings where γ≥1/√n. A stronger bound was recently proved by the authors (Feldman and Vondrak, 2018) that reduces the overhead to at most O(n1/4). Still, both of these results give optimal generalization bounds only when γ=O(1/n). We prove a nearly tight bound of O(γlog(n)log(n/δ)+√log(1/δ)/n) on the estimation error of any γ-uniformly stable algorithm. It implies that for algorithms that are uniformly stable with γ=O(1/√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.

Cite this Paper

BibTeX
@InProceedings{pmlr-v99-feldman19a, title = {High probability generalization bounds for uniformly stable algorithms with nearly optimal rate}, author = {Feldman, Vitaly and Vondrak, Jan}, booktitle = {Proceedings of the Thirty-Second Conference on Learning Theory}, pages = {1270--1279}, year = {2019}, editor = {Beygelzimer, Alina and Hsu, Daniel}, volume = {99}, series = {Proceedings of Machine Learning Research}, month = {25--28 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v99/feldman19a/feldman19a.pdf}, url = {https://proceedings.mlr.press/v99/feldman19a.html}, 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 (2001) 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.} }
Endnote
%0 Conference Paper %T High probability generalization bounds for uniformly stable algorithms with nearly optimal rate %A Vitaly Feldman %A Jan Vondrak %B Proceedings of the Thirty-Second Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2019 %E Alina Beygelzimer %E Daniel Hsu %F pmlr-v99-feldman19a %I PMLR %P 1270--1279 %U https://proceedings.mlr.press/v99/feldman19a.html %V 99 %X 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 (2001) 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.
APA
Feldman, V. & Vondrak, J.. (2019). High probability generalization bounds for uniformly stable algorithms with nearly optimal rate. Proceedings of the Thirty-Second Conference on Learning Theory, in Proceedings of Machine Learning Research 99:1270-1279 Available from https://proceedings.mlr.press/v99/feldman19a.html.

Related Material