Pairwise learning to rank by neural networks revisited: reconstruction, theoretical analysis and practical performance

Marius Köppel; Alexander Segner; Martin Wagener; Lukas Pensel; Andreas Karwath; Stefan Kramer

Pairwise learning to rank by neural networks revisited: reconstruction, theoretical analysis and practical performance

Marius Köppel, Alexander Segner, Martin Wagener, Lukas Pensel, Andreas Karwath, Stefan Kramer

Research output: Contribution to conference (unpublished) › Paper › peer-review

184 Downloads (Pure)

Abstract

We present a pairwise learning to rank approach based on a neural net, called DirectRanker, that generalizes the RankNet architecture. We show mathematically that our model is reflexive, antisymmetric, and transitive allowing for simplified training and improved performance. Experimental results on the LETOR MSLR-WEB10K, MQ2007 and MQ2008 datasets show that our model outperforms numerous state-of- the-art methods, while being inherently simpler in structure and using a pairwise approach only.

Original language	English
Publication status	Published - 16 Sept 2019
Event	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases - Würzburg, Germany, Würzburg, Germany Duration: 16 Sept 2019 → 20 Sept 2019 https://ecmlpkdd2019.org

Conference

Conference	European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases
Abbreviated title	ECMLPKDD
Country/Territory	Germany
City	Würzburg
Period	16/09/19 → 20/09/19
Internet address	https://ecmlpkdd2019.org

Keywords

Information Retrieval
Machine Learning
Learning to Rank

ASJC Scopus subject areas

Artificial Intelligence
Information Systems

Access to Document

400.pdfFinal published version, 1.04 MB

Andreas Karwath - Semantic-based secure infrastructure for interoperable, in-depth biomedical data and analytics
Cazier, J., Karwath, A., Yau, C. & Gkoutos, G.
Medical Research Council
14/02/18 → 13/02/21
Project: Research Councils

Cite this

@conference{9a176ab0a23b49d6b28b2afedab65737,

title = "Pairwise learning to rank by neural networks revisited: reconstruction, theoretical analysis and practical performance",

abstract = "We present a pairwise learning to rank approach based on a neural net, called DirectRanker, that generalizes the RankNet architecture. We show mathematically that our model is reflexive, antisymmetric, and transitive allowing for simplified training and improved performance. Experimental results on the LETOR MSLR-WEB10K, MQ2007 and MQ2008 datasets show that our model outperforms numerous state-of- the-art methods, while being inherently simpler in structure and using a pairwise approach only.",

keywords = "Information Retrieval, Machine Learning, Learning to Rank",

author = "Marius K{\"o}ppel and Alexander Segner and Martin Wagener and Lukas Pensel and Andreas Karwath and Stefan Kramer",

year = "2019",

month = sep,

day = "16",

language = "English",

note = "European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECMLPKDD ; Conference date: 16-09-2019 Through 20-09-2019",

url = "https://ecmlpkdd2019.org",

}

Pairwise learning to rank by neural networks revisited: reconstruction, theoretical analysis and practical performance. / Köppel, Marius; Segner, Alexander; Wagener, Martin et al.
2019. Paper presented at European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, Würzburg, Bavaria, Germany.

Research output: Contribution to conference (unpublished) › Paper › peer-review

TY - CONF

T1 - Pairwise learning to rank by neural networks revisited

T2 - European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases

AU - Köppel, Marius

AU - Segner, Alexander

AU - Wagener, Martin

AU - Pensel, Lukas

AU - Karwath, Andreas

AU - Kramer, Stefan

PY - 2019/9/16

Y1 - 2019/9/16

N2 - We present a pairwise learning to rank approach based on a neural net, called DirectRanker, that generalizes the RankNet architecture. We show mathematically that our model is reflexive, antisymmetric, and transitive allowing for simplified training and improved performance. Experimental results on the LETOR MSLR-WEB10K, MQ2007 and MQ2008 datasets show that our model outperforms numerous state-of- the-art methods, while being inherently simpler in structure and using a pairwise approach only.

AB - We present a pairwise learning to rank approach based on a neural net, called DirectRanker, that generalizes the RankNet architecture. We show mathematically that our model is reflexive, antisymmetric, and transitive allowing for simplified training and improved performance. Experimental results on the LETOR MSLR-WEB10K, MQ2007 and MQ2008 datasets show that our model outperforms numerous state-of- the-art methods, while being inherently simpler in structure and using a pairwise approach only.

KW - Information Retrieval

KW - Machine Learning

KW - Learning to Rank

UR - https://ecmlpkdd2019.org/downloads/paper/400.pdf

M3 - Paper

Y2 - 16 September 2019 through 20 September 2019

ER -

Pairwise learning to rank by neural networks revisited: reconstruction, theoretical analysis and practical performance

Abstract

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Projects

Andreas Karwath - Semantic-based secure infrastructure for interoperable, in-depth biomedical data and analytics

Cite this