Variational Inference from Ranked Samples with Features

Yuan Guo; Jennifer Dy; Deniz Erdoğmuş; Jayashree Kalpathy-Cramer; Susan Ostmo; J. Peter Campbell; Michael F. Chiang; Stratis Ioannidis

Variational Inference from Ranked Samples with Features

Yuan Guo, Jennifer Dy, Deniz Erdoğmuş, Jayashree Kalpathy-Cramer, Susan Ostmo, J. Peter Campbell, Michael F. Chiang, Stratis Ioannidis

Ophthalmology

Research output: Contribution to journal › Conference article › peer-review

1 Scopus citations

Abstract

In many supervised learning settings, elicited labels comprise pairwise comparisons or rankings of samples. We propose a Bayesian inference model for ranking datasets, allowing us to take a probabilistic approach to ranking inference. Our probabilistic assumptions are motivated by, and consistent with, the so-called Plackett-Luce model. We propose a variational inference method to extract a closed-form Gaussian posterior distribution. We show experimentally that the resulting posterior yields more reliable ranking predictions compared to predictions via point estimates.

Original language	English (US)
Pages (from-to)	599-614
Number of pages	16
Journal	Proceedings of Machine Learning Research
Volume	101
State	Published - 2019
Event	11th Asian Conference on Machine Learning, ACML 2019 - Nagoya, Japan Duration: Nov 17 2019 → Nov 19 2019

Keywords

Plackett Luce
Softmax Bound
Variational inference

ASJC Scopus subject areas

Artificial Intelligence
Software
Control and Systems Engineering
Statistics and Probability

Cite this

@article{30f4af039c3a45babf10e4310fe91569,

title = "Variational Inference from Ranked Samples with Features",

abstract = "In many supervised learning settings, elicited labels comprise pairwise comparisons or rankings of samples. We propose a Bayesian inference model for ranking datasets, allowing us to take a probabilistic approach to ranking inference. Our probabilistic assumptions are motivated by, and consistent with, the so-called Plackett-Luce model. We propose a variational inference method to extract a closed-form Gaussian posterior distribution. We show experimentally that the resulting posterior yields more reliable ranking predictions compared to predictions via point estimates.",

keywords = "Plackett Luce, Softmax Bound, Variational inference",

author = "Yuan Guo and Jennifer Dy and Deniz Erdoğmu{\c s} and Jayashree Kalpathy-Cramer and Susan Ostmo and Campbell, {J. Peter} and Chiang, {Michael F.} and Stratis Ioannidis",

note = "Publisher Copyright: {\textcopyright} 2019 Y. Guo1, J. Dy1, D. Erdogmu{\c s}1, J. Kalpathy-Cramer2, S. Ostmo3, J.P. Campbell3, M.F. Chiang3 & S. Ioannidis1.; 11th Asian Conference on Machine Learning, ACML 2019 ; Conference date: 17-11-2019 Through 19-11-2019",

year = "2019",

language = "English (US)",

volume = "101",

pages = "599--614",

journal = "Proceedings of Machine Learning Research",

issn = "2640-3498",

}

TY - JOUR

T1 - Variational Inference from Ranked Samples with Features

AU - Guo, Yuan

AU - Dy, Jennifer

AU - Erdoğmuş, Deniz

AU - Kalpathy-Cramer, Jayashree

AU - Ostmo, Susan

AU - Campbell, J. Peter

AU - Chiang, Michael F.

AU - Ioannidis, Stratis

PY - 2019

Y1 - 2019

N2 - In many supervised learning settings, elicited labels comprise pairwise comparisons or rankings of samples. We propose a Bayesian inference model for ranking datasets, allowing us to take a probabilistic approach to ranking inference. Our probabilistic assumptions are motivated by, and consistent with, the so-called Plackett-Luce model. We propose a variational inference method to extract a closed-form Gaussian posterior distribution. We show experimentally that the resulting posterior yields more reliable ranking predictions compared to predictions via point estimates.

AB - In many supervised learning settings, elicited labels comprise pairwise comparisons or rankings of samples. We propose a Bayesian inference model for ranking datasets, allowing us to take a probabilistic approach to ranking inference. Our probabilistic assumptions are motivated by, and consistent with, the so-called Plackett-Luce model. We propose a variational inference method to extract a closed-form Gaussian posterior distribution. We show experimentally that the resulting posterior yields more reliable ranking predictions compared to predictions via point estimates.

KW - Plackett Luce

KW - Softmax Bound

KW - Variational inference

UR - http://www.scopus.com/inward/record.url?scp=85107793433&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85107793433&partnerID=8YFLogxK

M3 - Conference article

AN - SCOPUS:85107793433

SN - 2640-3498

VL - 101

SP - 599

EP - 614

JO - Proceedings of Machine Learning Research

JF - Proceedings of Machine Learning Research

T2 - 11th Asian Conference on Machine Learning, ACML 2019

Y2 - 17 November 2019 through 19 November 2019

ER -

Variational Inference from Ranked Samples with Features

Abstract

Keywords

ASJC Scopus subject areas

Other files and links

Fingerprint

Cite this