The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning

Jesse H. Krijthe; Marco Loog

The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning

Pattern Recognition and Bioinformatics

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

4 Citations (Scopus)

Abstract

Consider a classification problem where we have both labeled and unlabeled data available. We show that for linear classifiers defined by convex margin-based surrogate losses that are decreasing, it is impossible to construct any semi-supervised approach that is able to guarantee an improvement over the supervised classifier measured by this surrogate loss on the labeled and unlabeled data. For convex margin-based loss functions that also increase, we demonstrate safe improvements are possible

Original language	English
Title of host publication	NIPS'18
Subtitle of host publication	Proceedings of the 32nd International Conference on Neural Information Processing Systems
Editors	S. Bengio, H.M. Wallach, H. Larochelle, K. Grauman, N. Cesa-Bianchi
Publisher	Curran Associates, Inc.
Pages	1793-1802
Number of pages	10
Publication status	Published - 2018
Event	NIPS 2018: 32nd Conference on Neural Information Processing Systems - Montréal, Canada Duration: 3 Dec 2018 → 8 Dec 2018 Conference number: 32

Conference

Conference	NIPS 2018
Country/Territory	Canada
City	Montréal
Period	3/12/18 → 8/12/18

Access to Document

https://dl.acm.org/citation.cfm?id=3327108

Cite this

Krijthe, J. H., & Loog, M. (2018). The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning. In S. Bengio, H. M. Wallach, H. Larochelle, K. Grauman, & N. Cesa-Bianchi (Eds.), NIPS'18: Proceedings of the 32nd International Conference on Neural Information Processing Systems (pp. 1793-1802). Curran Associates, Inc.. https://dl.acm.org/citation.cfm?id=3327108

@inproceedings{c01bedcfe42648c5880dcd9d74add688,

title = "The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning",

abstract = "Consider a classification problem where we have both labeled and unlabeled data available. We show that for linear classifiers defined by convex margin-based surrogate losses that are decreasing, it is impossible to construct any semi-supervised approach that is able to guarantee an improvement over the supervised classifier measured by this surrogate loss on the labeled and unlabeled data. For convex margin-based loss functions that also increase, we demonstrate safe improvements are possible",

author = "Krijthe, {Jesse H.} and Marco Loog",

year = "2018",

language = "English",

pages = "1793--1802",

editor = "S. Bengio and H.M. Wallach and H. Larochelle and K. Grauman and N. Cesa-Bianchi",

booktitle = "NIPS'18",

publisher = "Curran Associates, Inc.",

note = "NIPS 2018 : 32nd Conference on Neural Information Processing Systems ; Conference date: 03-12-2018 Through 08-12-2018",

}

Krijthe, JH & Loog, M 2018, The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning. in S Bengio, HM Wallach, H Larochelle, K Grauman & N Cesa-Bianchi (eds), NIPS'18: Proceedings of the 32nd International Conference on Neural Information Processing Systems. Curran Associates, Inc., pp. 1793-1802, NIPS 2018, Montréal, Canada, 3/12/18. <https://dl.acm.org/citation.cfm?id=3327108>

The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning. / Krijthe, Jesse H.; Loog, Marco.
NIPS'18: Proceedings of the 32nd International Conference on Neural Information Processing Systems. ed. / S. Bengio; H.M. Wallach; H. Larochelle; K. Grauman; N. Cesa-Bianchi. Curran Associates, Inc., 2018. p. 1793-1802.

Research output: Chapter in Book/Conference proceedings/Edited volume › Conference contribution › Scientific › peer-review

TY - GEN

T1 - The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning

AU - Krijthe, Jesse H.

AU - Loog, Marco

N1 - Conference code: 32

PY - 2018

Y1 - 2018

N2 - Consider a classification problem where we have both labeled and unlabeled data available. We show that for linear classifiers defined by convex margin-based surrogate losses that are decreasing, it is impossible to construct any semi-supervised approach that is able to guarantee an improvement over the supervised classifier measured by this surrogate loss on the labeled and unlabeled data. For convex margin-based loss functions that also increase, we demonstrate safe improvements are possible

AB - Consider a classification problem where we have both labeled and unlabeled data available. We show that for linear classifiers defined by convex margin-based surrogate losses that are decreasing, it is impossible to construct any semi-supervised approach that is able to guarantee an improvement over the supervised classifier measured by this surrogate loss on the labeled and unlabeled data. For convex margin-based loss functions that also increase, we demonstrate safe improvements are possible

M3 - Conference contribution

SP - 1793

EP - 1802

BT - NIPS'18

A2 - Bengio, S.

A2 - Wallach, H.M.

A2 - Larochelle, H.

A2 - Grauman, K.

A2 - Cesa-Bianchi, N.

PB - Curran Associates, Inc.

T2 - NIPS 2018

Y2 - 3 December 2018 through 8 December 2018

ER -

The Pessimistic Limits and Possibilities of Margin-based Losses in Semi-supervised Learning

Abstract

Conference

Access to Document

Fingerprint

Cite this