Minimax statistical learning with Wasserstein distances

Jaeho Lee, Maxim Raginsky

Research output: Contribution to journalConference article

Abstract

As opposed to standard empirical risk minimization (ERM), distributionally robust optimization aims to minimize the worst-case risk over a larger ambiguity set containing the original empirical distribution of the training data. In this work, we describe a minimax framework for statistical learning with ambiguity sets given by balls in Wasserstein space. In particular, we prove generalization bounds that involve the covering number properties of the original ERM problem. As an illustrative example, we provide generalization guarantees for transport-based domain adaptation problems where the Wasserstein distance between the source and target domain distributions can be reliably estimated from unlabeled samples.

Original languageEnglish (US)
Pages (from-to)2687-2696
Number of pages10
JournalAdvances in Neural Information Processing Systems
Volume2018-December
StatePublished - Jan 1 2018
Event32nd Conference on Neural Information Processing Systems, NeurIPS 2018 - Montreal, Canada
Duration: Dec 2 2018Dec 8 2018

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Minimax statistical learning with Wasserstein distances. / Lee, Jaeho; Raginsky, Maxim.

In: Advances in Neural Information Processing Systems, Vol. 2018-December, 01.01.2018, p. 2687-2696.

Research output: Contribution to journalConference article

@article{40f5f726ddc74e659aa10cbffd74984b,
title = "Minimax statistical learning with Wasserstein distances",
abstract = "As opposed to standard empirical risk minimization (ERM), distributionally robust optimization aims to minimize the worst-case risk over a larger ambiguity set containing the original empirical distribution of the training data. In this work, we describe a minimax framework for statistical learning with ambiguity sets given by balls in Wasserstein space. In particular, we prove generalization bounds that involve the covering number properties of the original ERM problem. As an illustrative example, we provide generalization guarantees for transport-based domain adaptation problems where the Wasserstein distance between the source and target domain distributions can be reliably estimated from unlabeled samples.",
author = "Jaeho Lee and Maxim Raginsky",
year = "2018",
month = "1",
day = "1",
language = "English (US)",
volume = "2018-December",
pages = "2687--2696",
journal = "Advances in Neural Information Processing Systems",
issn = "1049-5258",

}

TY - JOUR

T1 - Minimax statistical learning with Wasserstein distances

AU - Lee, Jaeho

AU - Raginsky, Maxim

PY - 2018/1/1

Y1 - 2018/1/1

N2 - As opposed to standard empirical risk minimization (ERM), distributionally robust optimization aims to minimize the worst-case risk over a larger ambiguity set containing the original empirical distribution of the training data. In this work, we describe a minimax framework for statistical learning with ambiguity sets given by balls in Wasserstein space. In particular, we prove generalization bounds that involve the covering number properties of the original ERM problem. As an illustrative example, we provide generalization guarantees for transport-based domain adaptation problems where the Wasserstein distance between the source and target domain distributions can be reliably estimated from unlabeled samples.

AB - As opposed to standard empirical risk minimization (ERM), distributionally robust optimization aims to minimize the worst-case risk over a larger ambiguity set containing the original empirical distribution of the training data. In this work, we describe a minimax framework for statistical learning with ambiguity sets given by balls in Wasserstein space. In particular, we prove generalization bounds that involve the covering number properties of the original ERM problem. As an illustrative example, we provide generalization guarantees for transport-based domain adaptation problems where the Wasserstein distance between the source and target domain distributions can be reliably estimated from unlabeled samples.

UR - http://www.scopus.com/inward/record.url?scp=85064814555&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85064814555&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:85064814555

VL - 2018-December

SP - 2687

EP - 2696

JO - Advances in Neural Information Processing Systems

JF - Advances in Neural Information Processing Systems

SN - 1049-5258

ER -