How hard is inference for structured prediction?

Amir Globerson, Tim Roughgarden, David Sontag, Cafer Yildirim

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Structured prediction tasks in machine learning involve the simultaneous prediction of multiple labels. This is often done by maximizing a score function on the space of labels, which decomposes as a sum of pairwise elements, each depending on two specific labels. The goal of this paper is to develop a theoretical explanation of the empirical effectiveness of heuristic inference algorithms for solving such structured prediction problems. We study the minimum-achievable expected Hamming error in such problems, highlighting the case of 2D grid graphs, which are common in machine vision applications. Our main theorems provide tight upper and lower bounds on this error, as well as a polynomial-time algorithm that achieves the bound.

Original languageEnglish (US)
Title of host publication32nd International Conference on Machine Learning, ICML 2015
PublisherInternational Machine Learning Society (IMLS)
Pages2171-2180
Number of pages10
Volume3
ISBN (Print)9781510810587
StatePublished - 2015
Event32nd International Conference on Machine Learning, ICML 2015 - Lile, France
Duration: Jul 6 2015Jul 11 2015

Other

Other32nd International Conference on Machine Learning, ICML 2015
CountryFrance
CityLile
Period7/6/157/11/15

Fingerprint

Labels
Computer vision
Learning systems
Polynomials

ASJC Scopus subject areas

  • Human-Computer Interaction
  • Computer Science Applications

Cite this

Globerson, A., Roughgarden, T., Sontag, D., & Yildirim, C. (2015). How hard is inference for structured prediction? In 32nd International Conference on Machine Learning, ICML 2015 (Vol. 3, pp. 2171-2180). International Machine Learning Society (IMLS).

How hard is inference for structured prediction? / Globerson, Amir; Roughgarden, Tim; Sontag, David; Yildirim, Cafer.

32nd International Conference on Machine Learning, ICML 2015. Vol. 3 International Machine Learning Society (IMLS), 2015. p. 2171-2180.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Globerson, A, Roughgarden, T, Sontag, D & Yildirim, C 2015, How hard is inference for structured prediction? in 32nd International Conference on Machine Learning, ICML 2015. vol. 3, International Machine Learning Society (IMLS), pp. 2171-2180, 32nd International Conference on Machine Learning, ICML 2015, Lile, France, 7/6/15.
Globerson A, Roughgarden T, Sontag D, Yildirim C. How hard is inference for structured prediction? In 32nd International Conference on Machine Learning, ICML 2015. Vol. 3. International Machine Learning Society (IMLS). 2015. p. 2171-2180
Globerson, Amir ; Roughgarden, Tim ; Sontag, David ; Yildirim, Cafer. / How hard is inference for structured prediction?. 32nd International Conference on Machine Learning, ICML 2015. Vol. 3 International Machine Learning Society (IMLS), 2015. pp. 2171-2180
@inproceedings{c88676912af245a39e8c703bcea71c67,
title = "How hard is inference for structured prediction?",
abstract = "Structured prediction tasks in machine learning involve the simultaneous prediction of multiple labels. This is often done by maximizing a score function on the space of labels, which decomposes as a sum of pairwise elements, each depending on two specific labels. The goal of this paper is to develop a theoretical explanation of the empirical effectiveness of heuristic inference algorithms for solving such structured prediction problems. We study the minimum-achievable expected Hamming error in such problems, highlighting the case of 2D grid graphs, which are common in machine vision applications. Our main theorems provide tight upper and lower bounds on this error, as well as a polynomial-time algorithm that achieves the bound.",
author = "Amir Globerson and Tim Roughgarden and David Sontag and Cafer Yildirim",
year = "2015",
language = "English (US)",
isbn = "9781510810587",
volume = "3",
pages = "2171--2180",
booktitle = "32nd International Conference on Machine Learning, ICML 2015",
publisher = "International Machine Learning Society (IMLS)",

}

TY - GEN

T1 - How hard is inference for structured prediction?

AU - Globerson, Amir

AU - Roughgarden, Tim

AU - Sontag, David

AU - Yildirim, Cafer

PY - 2015

Y1 - 2015

N2 - Structured prediction tasks in machine learning involve the simultaneous prediction of multiple labels. This is often done by maximizing a score function on the space of labels, which decomposes as a sum of pairwise elements, each depending on two specific labels. The goal of this paper is to develop a theoretical explanation of the empirical effectiveness of heuristic inference algorithms for solving such structured prediction problems. We study the minimum-achievable expected Hamming error in such problems, highlighting the case of 2D grid graphs, which are common in machine vision applications. Our main theorems provide tight upper and lower bounds on this error, as well as a polynomial-time algorithm that achieves the bound.

AB - Structured prediction tasks in machine learning involve the simultaneous prediction of multiple labels. This is often done by maximizing a score function on the space of labels, which decomposes as a sum of pairwise elements, each depending on two specific labels. The goal of this paper is to develop a theoretical explanation of the empirical effectiveness of heuristic inference algorithms for solving such structured prediction problems. We study the minimum-achievable expected Hamming error in such problems, highlighting the case of 2D grid graphs, which are common in machine vision applications. Our main theorems provide tight upper and lower bounds on this error, as well as a polynomial-time algorithm that achieves the bound.

UR - http://www.scopus.com/inward/record.url?scp=84969961954&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84969961954&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9781510810587

VL - 3

SP - 2171

EP - 2180

BT - 32nd International Conference on Machine Learning, ICML 2015

PB - International Machine Learning Society (IMLS)

ER -