Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios

Miquel Perelló-Nieto; Raúl Santos-Rodríguez; Jesús Cid-Sueiro

doi:10.1007/978-3-319-68765-0_21

Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios

Miquel Perelló-Nieto^*, Raúl Santos-Rodríguez, Jesús Cid-Sueiro

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

4 Citations (Scopus)

373 Downloads (Pure)

Abstract

In many real-world problems, labels are often weak, meaning that each instance is labelled as belonging to one of several candidate categories, at most one of them being true. Recent theoretical contributions have shown that it is possible to construct proper losses or classification calibrated losses for weakly labelled classification scenarios by means of a linear transformation of conventional proper or classification calibrated losses, respectively. However, how to translate these theoretical results into practice has not been explored yet. This paper discusses both the algorithmic design and the potential advantages of this approach, analyzing consistency and convexity issues arising in practical settings, and evaluating the behavior of such transformations under different types of weak labels.

Original language	English
Title of host publication	Advances in Intelligent Data Analysis XVI
Subtitle of host publication	16th International Symposium, IDA 2017, London, UK, October 26–28, 2017, Proceedings
Publisher	Springer Verlag
Pages	247-259
Number of pages	13
ISBN (Electronic)	9783319687650
ISBN (Print)	9783319687643
DOIs	https://doi.org/10.1007/978-3-319-68765-0_21
Publication status	Published - 4 Oct 2017
Event	16th International Symposium on Intelligent Data Analysis, IDA 2017 - London, United Kingdom Duration: 26 Oct 2017 → 28 Oct 2017

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10584 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	16th International Symposium on Intelligent Data Analysis, IDA 2017
Country/Territory	United Kingdom
City	London
Period	26/10/17 → 28/10/17

Research Groups and Themes

Digital Health
SPHERE

Keywords

Noisy labels
Proper losses
Weak labels

Access to Document

10.1007/978-3-319-68765-0_21

Full-text PDF (accepted author manuscript)
This is the author accepted manuscript (AAM). The final published version (version of record) is available online via Springer at https://link.springer.com/chapter/10.1007%2F978-3-319-68765-0_21 . Please refer to any applicable terms of use of the publisher.
Accepted author manuscript, 740 KB

Handle.net

Persistent link

Cite this

Perelló-Nieto, M., Santos-Rodríguez, R., & Cid-Sueiro, J. (2017). Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios. In Advances in Intelligent Data Analysis XVI: 16th International Symposium, IDA 2017, London, UK, October 26–28, 2017, Proceedings (pp. 247-259). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10584 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-68765-0_21

Perelló-Nieto, Miquel ; Santos-Rodríguez, Raúl ; Cid-Sueiro, Jesús. / Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios. Advances in Intelligent Data Analysis XVI: 16th International Symposium, IDA 2017, London, UK, October 26–28, 2017, Proceedings. Springer Verlag, 2017. pp. 247-259 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{1ce530b5c552416ea961d9dcbe7f902e,

title = "Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios",

abstract = "In many real-world problems, labels are often weak, meaning that each instance is labelled as belonging to one of several candidate categories, at most one of them being true. Recent theoretical contributions have shown that it is possible to construct proper losses or classification calibrated losses for weakly labelled classification scenarios by means of a linear transformation of conventional proper or classification calibrated losses, respectively. However, how to translate these theoretical results into practice has not been explored yet. This paper discusses both the algorithmic design and the potential advantages of this approach, analyzing consistency and convexity issues arising in practical settings, and evaluating the behavior of such transformations under different types of weak labels.",

keywords = "Noisy labels, Proper losses, Weak labels",

author = "Miquel Perell{\'o}-Nieto and Ra{\'u}l Santos-Rodr{\'i}guez and Jes{\'u}s Cid-Sueiro",

year = "2017",

month = oct,

day = "4",

doi = "10.1007/978-3-319-68765-0_21",

language = "English",

isbn = "9783319687643",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "247--259",

booktitle = "Advances in Intelligent Data Analysis XVI",

address = "Germany",

note = "16th International Symposium on Intelligent Data Analysis, IDA 2017 ; Conference date: 26-10-2017 Through 28-10-2017",

}

Perelló-Nieto, M , Santos-Rodríguez, R & Cid-Sueiro, J 2017, Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios. in Advances in Intelligent Data Analysis XVI: 16th International Symposium, IDA 2017, London, UK, October 26–28, 2017, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10584 LNCS, Springer Verlag, pp. 247-259, 16th International Symposium on Intelligent Data Analysis, IDA 2017, London, United Kingdom, 26/10/17. https://doi.org/10.1007/978-3-319-68765-0_21

Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios. / Perelló-Nieto, Miquel ; Santos-Rodríguez, Raúl; Cid-Sueiro, Jesús.
Advances in Intelligent Data Analysis XVI: 16th International Symposium, IDA 2017, London, UK, October 26–28, 2017, Proceedings. Springer Verlag, 2017. p. 247-259 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10584 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference Contribution (Conference Proceeding)

TY - GEN

T1 - Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios

AU - Perelló-Nieto, Miquel

AU - Santos-Rodríguez, Raúl

AU - Cid-Sueiro, Jesús

PY - 2017/10/4

Y1 - 2017/10/4

N2 - In many real-world problems, labels are often weak, meaning that each instance is labelled as belonging to one of several candidate categories, at most one of them being true. Recent theoretical contributions have shown that it is possible to construct proper losses or classification calibrated losses for weakly labelled classification scenarios by means of a linear transformation of conventional proper or classification calibrated losses, respectively. However, how to translate these theoretical results into practice has not been explored yet. This paper discusses both the algorithmic design and the potential advantages of this approach, analyzing consistency and convexity issues arising in practical settings, and evaluating the behavior of such transformations under different types of weak labels.

AB - In many real-world problems, labels are often weak, meaning that each instance is labelled as belonging to one of several candidate categories, at most one of them being true. Recent theoretical contributions have shown that it is possible to construct proper losses or classification calibrated losses for weakly labelled classification scenarios by means of a linear transformation of conventional proper or classification calibrated losses, respectively. However, how to translate these theoretical results into practice has not been explored yet. This paper discusses both the algorithmic design and the potential advantages of this approach, analyzing consistency and convexity issues arising in practical settings, and evaluating the behavior of such transformations under different types of weak labels.

KW - Noisy labels

KW - Proper losses

KW - Weak labels

UR - http://www.scopus.com/inward/record.url?scp=85032653048&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-68765-0_21

DO - 10.1007/978-3-319-68765-0_21

M3 - Conference Contribution (Conference Proceeding)

AN - SCOPUS:85032653048

SN - 9783319687643

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 247

EP - 259

BT - Advances in Intelligent Data Analysis XVI

PB - Springer Verlag

T2 - 16th International Symposium on Intelligent Data Analysis, IDA 2017

Y2 - 26 October 2017 through 28 October 2017

ER -

Perelló-Nieto M , Santos-Rodríguez R, Cid-Sueiro J. Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios. In Advances in Intelligent Data Analysis XVI: 16th International Symposium, IDA 2017, London, UK, October 26–28, 2017, Proceedings. Springer Verlag. 2017. p. 247-259. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-68765-0_21

Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios

Abstract

Publication series

Conference

Research Groups and Themes

Keywords

Access to Document

Handle.net

Other files and links

Fingerprint

8032 EPSRC INSTITUTIONAL SPONSORSHIP

SPHERE (EPSRC IRC)

Uncertainty aware classification: augmenting classifiers to handle uncertainty

HPC (High Performance Computing) and HTC (High Throughput Computing) Facilities

Cite this

Adapting Supervised Classification Algorithms to Arbitrary Weak Label Scenarios

Abstract

Publication series

Conference

Research Groups and Themes

Keywords

Access to Document

Handle.net

Other files and links

Fingerprint

Projects

8032 EPSRC INSTITUTIONAL SPONSORSHIP

SPHERE (EPSRC IRC)

Student theses

Uncertainty aware classification: augmenting classifiers to handle uncertainty

Equipment

HPC (High Performance Computing) and HTC (High Throughput Computing) Facilities

Cite this