Latent supervised learning

Susan Wei; Michael R. Kosorok

doi:10.1080/01621459.2013.789695

Latent supervised learning

Susan Wei, Michael R. Kosorok

Research output: Contribution to journal › Article › peer-review

20 Scopus citations

Abstract

This article introduces a new machine learning task, called latent supervised learning, where the goal is to learn a binary classifier from continuous training labels that serve as surrogates for the unobserved class labels. We investigate a specific model where the surrogate variable arises from a two-component Gaussian mixture with unknown means and variances, and the component membership is determined by a hyperplane in the covariate space. The estimation of the separating hyperplane and the Gaussian mixture parameters forms what shall be referred to as the change-line classification problem. We propose a data-driven sieve maximum likelihood estimator for the hyperplane, which in turn can be used to estimate the parameters of the Gaussian mixture. The estimator is shown to be consistent. Simulations as well as empirical data show the estimator has high classification accuracy.

Original language	English (US)
Pages (from-to)	957-970
Number of pages	14
Journal	Journal of the American Statistical Association
Volume	108
Issue number	503
DOIs	https://doi.org/10.1080/01621459.2013.789695
State	Published - 2013
Externally published	Yes

Bibliographical note

Funding Information:
Susan Wei is Doctoral Student, Department of Statistics and Operations Research, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 (Email: susanwe@live.unc.edu). Michael R. Kosorok is Professor and Chair, Department of Biostatistics and Professor, Department of Statistics and Operations Research, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 (Email: kosorok@unc.edu). The first author was funded through the National Science Foundation Graduate Fellowship and the National Institutes of Health (NIH) grant T32 GM067553-05S1. The second author was funded in part by the NIH grant CA142538. We thank Editor Xuming He, the Associate Editor, and two anonymous referees for their helpful comments that led to a significantly improved article.

Keywords

Classification and clustering
Glivenko-Cantelli classes
Sieve maximum likelihood estimation
Sliced inverse regression
Statistical learning

Access

10.1080/01621459.2013.789695

OpenUrl availability

Full text

Cite this

@article{34b686b3d26f477e931088026d183f19,

title = "Latent supervised learning",

abstract = "This article introduces a new machine learning task, called latent supervised learning, where the goal is to learn a binary classifier from continuous training labels that serve as surrogates for the unobserved class labels. We investigate a specific model where the surrogate variable arises from a two-component Gaussian mixture with unknown means and variances, and the component membership is determined by a hyperplane in the covariate space. The estimation of the separating hyperplane and the Gaussian mixture parameters forms what shall be referred to as the change-line classification problem. We propose a data-driven sieve maximum likelihood estimator for the hyperplane, which in turn can be used to estimate the parameters of the Gaussian mixture. The estimator is shown to be consistent. Simulations as well as empirical data show the estimator has high classification accuracy.",

keywords = "Classification and clustering, Glivenko-Cantelli classes, Sieve maximum likelihood estimation, Sliced inverse regression, Statistical learning",

author = "Susan Wei and Kosorok, {Michael R.}",

note = "Funding Information: Susan Wei is Doctoral Student, Department of Statistics and Operations Research, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 (Email: susanwe@live.unc.edu). Michael R. Kosorok is Professor and Chair, Department of Biostatistics and Professor, Department of Statistics and Operations Research, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 (Email: kosorok@unc.edu). The first author was funded through the National Science Foundation Graduate Fellowship and the National Institutes of Health (NIH) grant T32 GM067553-05S1. The second author was funded in part by the NIH grant CA142538. We thank Editor Xuming He, the Associate Editor, and two anonymous referees for their helpful comments that led to a significantly improved article.",

year = "2013",

doi = "10.1080/01621459.2013.789695",

language = "English (US)",

volume = "108",

pages = "957--970",

journal = "Journal of the American Statistical Association",

issn = "0162-1459",

publisher = "Taylor and Francis Ltd.",

number = "503",

}

TY - JOUR

T1 - Latent supervised learning

AU - Wei, Susan

AU - Kosorok, Michael R.

N1 - Funding Information: Susan Wei is Doctoral Student, Department of Statistics and Operations Research, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 (Email: susanwe@live.unc.edu). Michael R. Kosorok is Professor and Chair, Department of Biostatistics and Professor, Department of Statistics and Operations Research, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 (Email: kosorok@unc.edu). The first author was funded through the National Science Foundation Graduate Fellowship and the National Institutes of Health (NIH) grant T32 GM067553-05S1. The second author was funded in part by the NIH grant CA142538. We thank Editor Xuming He, the Associate Editor, and two anonymous referees for their helpful comments that led to a significantly improved article.

PY - 2013

Y1 - 2013

N2 - This article introduces a new machine learning task, called latent supervised learning, where the goal is to learn a binary classifier from continuous training labels that serve as surrogates for the unobserved class labels. We investigate a specific model where the surrogate variable arises from a two-component Gaussian mixture with unknown means and variances, and the component membership is determined by a hyperplane in the covariate space. The estimation of the separating hyperplane and the Gaussian mixture parameters forms what shall be referred to as the change-line classification problem. We propose a data-driven sieve maximum likelihood estimator for the hyperplane, which in turn can be used to estimate the parameters of the Gaussian mixture. The estimator is shown to be consistent. Simulations as well as empirical data show the estimator has high classification accuracy.

AB - This article introduces a new machine learning task, called latent supervised learning, where the goal is to learn a binary classifier from continuous training labels that serve as surrogates for the unobserved class labels. We investigate a specific model where the surrogate variable arises from a two-component Gaussian mixture with unknown means and variances, and the component membership is determined by a hyperplane in the covariate space. The estimation of the separating hyperplane and the Gaussian mixture parameters forms what shall be referred to as the change-line classification problem. We propose a data-driven sieve maximum likelihood estimator for the hyperplane, which in turn can be used to estimate the parameters of the Gaussian mixture. The estimator is shown to be consistent. Simulations as well as empirical data show the estimator has high classification accuracy.

KW - Classification and clustering

KW - Glivenko-Cantelli classes

KW - Sieve maximum likelihood estimation

KW - Sliced inverse regression

KW - Statistical learning

UR - http://www.scopus.com/inward/record.url?scp=84890035853&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84890035853&partnerID=8YFLogxK

U2 - 10.1080/01621459.2013.789695

DO - 10.1080/01621459.2013.789695

M3 - Article

C2 - 24319303

AN - SCOPUS:84890035853

SN - 0162-1459

VL - 108

SP - 957

EP - 970

JO - Journal of the American Statistical Association

JF - Journal of the American Statistical Association

IS - 503

ER -

Latent supervised learning

Abstract

Bibliographical note

Keywords

Access

OpenUrl availability

Other files and links

Fingerprint

Cite this