On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification

Erik Schultheis; Rohit Babbar; Marek Wydmuch; Krzysztof Dembczynski

doi:10.1145/3534678.3539466

On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification

Erik Schultheis, Rohit Babbar, Marek Wydmuch, Krzysztof Dembczynski

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

26 Citations (Scopus)

80 Downloads (Pure)

Abstract

The propensity model introduced by Jain et al has become a standard approach for dealing with missing and long-tail labels in extreme multi-label classification (XMLC). In this paper, we critically revise this approach showing that despite its theoretical soundness, its application in contemporary XMLC works is debatable. We exhaustively discuss the flaws of the propensity-based approach, and present several recipes, some of them related to solutions used in search engines and recommender systems, that we believe constitute promising alternatives to be followed in XMLC.

Original language	English
Title of host publication	Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining
Publisher	ACM
Pages	1547–1557
Number of pages	11
ISBN (Electronic)	978-1-4503-9385-0
DOIs	https://doi.org/10.1145/3534678.3539466
Publication status	Published - Aug 2022
MoE publication type	A4 Conference publication
Event	ACM SIGKDD International Conference on Knowledge Discovery and Data Mining - Washington, United States Duration: 14 Aug 2022 → 18 Aug 2022 Conference number: 28 https://kdd.org/kdd2022/

Conference

Conference	ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
Abbreviated title	KDD
Country/Territory	United States
City	Washington
Period	14/08/2022 → 18/08/2022
Internet address	https://kdd.org/kdd2022/

Access to Document

10.1145/3534678.3539466

3534678.3539466Final published version, 1.55 MB

Cite this

@inproceedings{28259d188a054073a16251902649b763,

title = "On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification",

abstract = "The propensity model introduced by Jain et al has become a standard approach for dealing with missing and long-tail labels in extreme multi-label classification (XMLC). In this paper, we critically revise this approach showing that despite its theoretical soundness, its application in contemporary XMLC works is debatable. We exhaustively discuss the flaws of the propensity-based approach, and present several recipes, some of them related to solutions used in search engines and recommender systems, that we believe constitute promising alternatives to be followed in XMLC.",

author = "Erik Schultheis and Rohit Babbar and Marek Wydmuch and Krzysztof Dembczynski",

year = "2022",

month = aug,

doi = "10.1145/3534678.3539466",

language = "English",

pages = "1547–1557",

booktitle = "Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining",

publisher = "ACM",

address = "United States",

note = "ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD ; Conference date: 14-08-2022 Through 18-08-2022",

url = "https://kdd.org/kdd2022/",

}

Schultheis, E, Babbar, R, Wydmuch, M & Dembczynski, K 2022, On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification. in Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. ACM, pp. 1547–1557, ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, Washington, District of Columbia, United States, 14/08/2022. https://doi.org/10.1145/3534678.3539466

On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification. / Schultheis, Erik; Babbar, Rohit; Wydmuch, Marek et al.
Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. ACM, 2022. p. 1547–1557.

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Scientific › peer-review

TY - GEN

T1 - On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification

AU - Schultheis, Erik

AU - Babbar, Rohit

AU - Wydmuch, Marek

AU - Dembczynski, Krzysztof

N1 - Conference code: 28

PY - 2022/8

Y1 - 2022/8

N2 - The propensity model introduced by Jain et al has become a standard approach for dealing with missing and long-tail labels in extreme multi-label classification (XMLC). In this paper, we critically revise this approach showing that despite its theoretical soundness, its application in contemporary XMLC works is debatable. We exhaustively discuss the flaws of the propensity-based approach, and present several recipes, some of them related to solutions used in search engines and recommender systems, that we believe constitute promising alternatives to be followed in XMLC.

AB - The propensity model introduced by Jain et al has become a standard approach for dealing with missing and long-tail labels in extreme multi-label classification (XMLC). In this paper, we critically revise this approach showing that despite its theoretical soundness, its application in contemporary XMLC works is debatable. We exhaustively discuss the flaws of the propensity-based approach, and present several recipes, some of them related to solutions used in search engines and recommender systems, that we believe constitute promising alternatives to be followed in XMLC.

UR - http://www.scopus.com/inward/record.url?scp=85137144550&partnerID=8YFLogxK

U2 - 10.1145/3534678.3539466

DO - 10.1145/3534678.3539466

M3 - Conference article in proceedings

SP - 1547

EP - 1557

BT - Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining

PB - ACM

T2 - ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

Y2 - 14 August 2022 through 18 August 2022

ER -

On Missing Labels, Long-tails and Propensities in Extreme Multi-label Classification

Abstract

Conference

Access to Document

Other files and links

Fingerprint

HPC-HD/Babbar: High Performance Computing for the Detection and Analysis of Historical Discourses