A Gaussian process mixture prior for hearing loss modeling

M.G.H. Cox; A. de Vries

A Gaussian process mixture prior for hearing loss modeling

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

172 Downloads (Pure)

Abstract

Machine learning approaches to hearing loss estimation can significantly reduce the number of required experiments, but require a good probabilistic hearing loss model. In this work we introduce such a model, obtained by fitting a mixture of Gaussian processes to a vast database containing audiometric records of around 85k people. The learned model can be used as a prior distribution for hearing loss, and can be conditioned on age and gender. Evaluation on a test set shows that our model outperforms an optimized Gaussian process model in terms of predictive accuracy.

Original language	English
Title of host publication	Benelearn 2017: Proceedings of the Twenty-Sixth Benelux Conference on Machine Learning, Technische Universiteit Eindhoven, 9-10 June 2017
Editors	W. Duivesteijn, M. Pechenizkiy, G.H.L. Fletcher
Pages	74-76
Publication status	Published - 9 Jun 2017
Event	Annual machine learning conference of the Benelux (Benelearn 2017) - Eindhoven, Netherlands Duration: 9 Jun 2017 → 10 Jun 2017 http://wwwis.win.tue.nl/~benelearn2017/

Conference

Conference	Annual machine learning conference of the Benelux (Benelearn 2017)
Abbreviated title	Benelearn 2017
Country/Territory	Netherlands
City	Eindhoven
Period	9/06/17 → 10/06/17
Internet address	http://wwwis.win.tue.nl/~benelearn2017/

Access to Document

publisher versionFinal published version, 805 KB

Cite this

@inproceedings{7745961c391343a8b7eb60d52c8be247,

title = "A Gaussian process mixture prior for hearing loss modeling",

abstract = "Machine learning approaches to hearing loss estimation can significantly reduce the number of required experiments, but require a good probabilistic hearing loss model. In this work we introduce such a model, obtained by fitting a mixture of Gaussian processes to a vast database containing audiometric records of around 85k people. The learned model can be used as a prior distribution for hearing loss, and can be conditioned on age and gender. Evaluation on a test set shows that our model outperforms an optimized Gaussian process model in terms of predictive accuracy.",

author = "M.G.H. Cox and {de Vries}, A.",

year = "2017",

month = jun,

day = "9",

language = "English",

pages = "74--76",

editor = "W. Duivesteijn and M. Pechenizkiy and {Fletcher }, G.H.L.",

booktitle = "Benelearn 2017: Proceedings of the Twenty-Sixth Benelux Conference on Machine Learning, Technische Universiteit Eindhoven, 9-10 June 2017",

note = "Annual machine learning conference of the Benelux (Benelearn 2017), Benelearn 2017 ; Conference date: 09-06-2017 Through 10-06-2017",

url = "http://wwwis.win.tue.nl/~benelearn2017/",

}

Cox, MGH & de Vries, A 2017, A Gaussian process mixture prior for hearing loss modeling. in W Duivesteijn, M Pechenizkiy & GHL Fletcher (eds), Benelearn 2017: Proceedings of the Twenty-Sixth Benelux Conference on Machine Learning, Technische Universiteit Eindhoven, 9-10 June 2017. pp. 74-76, Annual machine learning conference of the Benelux (Benelearn 2017), Eindhoven, Netherlands, 9/06/17.

A Gaussian process mixture prior for hearing loss modeling. / Cox, M.G.H.; de Vries, A.
Benelearn 2017: Proceedings of the Twenty-Sixth Benelux Conference on Machine Learning, Technische Universiteit Eindhoven, 9-10 June 2017. ed. / W. Duivesteijn; M. Pechenizkiy; G.H.L. Fletcher . 2017. p. 74-76.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - A Gaussian process mixture prior for hearing loss modeling

AU - Cox, M.G.H.

AU - de Vries, A.

PY - 2017/6/9

Y1 - 2017/6/9

N2 - Machine learning approaches to hearing loss estimation can significantly reduce the number of required experiments, but require a good probabilistic hearing loss model. In this work we introduce such a model, obtained by fitting a mixture of Gaussian processes to a vast database containing audiometric records of around 85k people. The learned model can be used as a prior distribution for hearing loss, and can be conditioned on age and gender. Evaluation on a test set shows that our model outperforms an optimized Gaussian process model in terms of predictive accuracy.

AB - Machine learning approaches to hearing loss estimation can significantly reduce the number of required experiments, but require a good probabilistic hearing loss model. In this work we introduce such a model, obtained by fitting a mixture of Gaussian processes to a vast database containing audiometric records of around 85k people. The learned model can be used as a prior distribution for hearing loss, and can be conditioned on age and gender. Evaluation on a test set shows that our model outperforms an optimized Gaussian process model in terms of predictive accuracy.

M3 - Conference contribution

SP - 74

EP - 76

BT - Benelearn 2017: Proceedings of the Twenty-Sixth Benelux Conference on Machine Learning, Technische Universiteit Eindhoven, 9-10 June 2017

A2 - Duivesteijn, W.

A2 - Pechenizkiy, M.

A2 - Fletcher , G.H.L.

T2 - Annual machine learning conference of the Benelux (Benelearn 2017)

Y2 - 9 June 2017 through 10 June 2017

ER -

A Gaussian process mixture prior for hearing loss modeling

Abstract

Conference

Access to Document

Fingerprint

Cite this