Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers

Sami Remes; Tommi Mononen; Samuel Kaski

Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers

Sami Remes, Tommi Mononen, Samuel Kaski

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Professional

Abstract

We propose a novel classification model for weak signal data, building upon a recent model for Bayesian multi-view learning, Group Factor Analysis (GFA). Instead of assuming all data to come from a single GFA model, we allow latent clusters, each having a different GFA model and producing a different class distribution. We show that sharing information across the clusters, by sharing factors, increases the classification accuracy considerably; the shared factors essentially form a flexible noise model that explains away the part of data not related to classification. Motivation for the setting comes from single-trial functional brain imaging data, having a very low signal-to-noise ratio and a natural multi-view setting, with the different sensors, measurement modalities (EEG, MEG, fMRI) and possible auxiliary information as views. We demonstrate our model on a MEG dataset.

Original language	English
Title of host publication	Proceedings of the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at NIPS 2015
Editors	Irina Rish, Leila Wehbe, Georg Langs, Moritz Grosse-Wentrup, Brian Murphy, Guillermo Cecchi
Place of Publication	https://arxiv.org/abs/1605.04435
Number of pages	8
Publication status	Published - 14 May 2016
MoE publication type	D3 Professional conference proceedings
Event	Workshop on Machine Learning and Interpretation in Neuroimaging - Montreal, Canada Duration: 11 Dec 2015 → 12 Dec 2015 Conference number: 5 https://sites.google.com/site/mliniworkshop2015/

Workshop

Workshop	Workshop on Machine Learning and Interpretation in Neuroimaging
Abbreviated title	MLINI
Country/Territory	Canada
City	Montreal
Period	11/12/2015 → 12/12/2015
Internet address	https://sites.google.com/site/mliniworkshop2015/

Keywords

Bayesian group factor analysis
brain decoding
MEG
multi-view learning
variational Bayesian inference

Access to Document

https://arxiv.org/abs/1512.05610

Cite this

Remes, S., Mononen, T., & Kaski, S. (2016). Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers. In I. Rish, L. Wehbe, G. Langs, M. Grosse-Wentrup, B. Murphy, & G. Cecchi (Eds.), Proceedings of the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at NIPS 2015 https://arxiv.org/abs/1512.05610

Remes, Sami ; Mononen, Tommi ; Kaski, Samuel. / Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers. Proceedings of the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at NIPS 2015. editor / Irina Rish ; Leila Wehbe ; Georg Langs ; Moritz Grosse-Wentrup ; Brian Murphy ; Guillermo Cecchi. https://arxiv.org/abs/1605.04435, 2016.

@inproceedings{7434908f458a4b6e9b530cd27ed5bfad,

title = "Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers",

abstract = "We propose a novel classification model for weak signal data, building upon a recent model for Bayesian multi-view learning, Group Factor Analysis (GFA). Instead of assuming all data to come from a single GFA model, we allow latent clusters, each having a different GFA model and producing a different class distribution. We show that sharing information across the clusters, by sharing factors, increases the classification accuracy considerably; the shared factors essentially form a flexible noise model that explains away the part of data not related to classification. Motivation for the setting comes from single-trial functional brain imaging data, having a very low signal-to-noise ratio and a natural multi-view setting, with the different sensors, measurement modalities (EEG, MEG, fMRI) and possible auxiliary information as views. We demonstrate our model on a MEG dataset.",

keywords = "Bayesian group factor analysis, brain decoding, MEG, multi-view learning, variational Bayesian inference",

author = "Sami Remes and Tommi Mononen and Samuel Kaski",

year = "2016",

month = may,

day = "14",

language = "English",

editor = "Irina Rish and Leila Wehbe and Georg Langs and Moritz Grosse-Wentrup and Brian Murphy and Guillermo Cecchi",

booktitle = "Proceedings of the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at NIPS 2015",

note = "Workshop on Machine Learning and Interpretation in Neuroimaging , MLINI ; Conference date: 11-12-2015 Through 12-12-2015",

url = "https://sites.google.com/site/mliniworkshop2015/",

}

Remes, S, Mononen, T & Kaski, S 2016, Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers. in I Rish, L Wehbe, G Langs, M Grosse-Wentrup, B Murphy & G Cecchi (eds), Proceedings of the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at NIPS 2015. https://arxiv.org/abs/1605.04435, Workshop on Machine Learning and Interpretation in Neuroimaging , Montreal, Canada, 11/12/2015. <https://arxiv.org/abs/1512.05610>

Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers. / Remes, Sami; Mononen, Tommi; Kaski, Samuel.
Proceedings of the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at NIPS 2015. ed. / Irina Rish; Leila Wehbe; Georg Langs; Moritz Grosse-Wentrup; Brian Murphy; Guillermo Cecchi. https://arxiv.org/abs/1605.04435, 2016.

Research output: Chapter in Book/Report/Conference proceeding › Conference article in proceedings › Professional

TY - GEN

T1 - Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers

AU - Remes, Sami

AU - Mononen, Tommi

AU - Kaski, Samuel

N1 - Conference code: 5

PY - 2016/5/14

Y1 - 2016/5/14

N2 - We propose a novel classification model for weak signal data, building upon a recent model for Bayesian multi-view learning, Group Factor Analysis (GFA). Instead of assuming all data to come from a single GFA model, we allow latent clusters, each having a different GFA model and producing a different class distribution. We show that sharing information across the clusters, by sharing factors, increases the classification accuracy considerably; the shared factors essentially form a flexible noise model that explains away the part of data not related to classification. Motivation for the setting comes from single-trial functional brain imaging data, having a very low signal-to-noise ratio and a natural multi-view setting, with the different sensors, measurement modalities (EEG, MEG, fMRI) and possible auxiliary information as views. We demonstrate our model on a MEG dataset.

AB - We propose a novel classification model for weak signal data, building upon a recent model for Bayesian multi-view learning, Group Factor Analysis (GFA). Instead of assuming all data to come from a single GFA model, we allow latent clusters, each having a different GFA model and producing a different class distribution. We show that sharing information across the clusters, by sharing factors, increases the classification accuracy considerably; the shared factors essentially form a flexible noise model that explains away the part of data not related to classification. Motivation for the setting comes from single-trial functional brain imaging data, having a very low signal-to-noise ratio and a natural multi-view setting, with the different sensors, measurement modalities (EEG, MEG, fMRI) and possible auxiliary information as views. We demonstrate our model on a MEG dataset.

KW - Bayesian group factor analysis

KW - brain decoding

KW - MEG

KW - multi-view learning

KW - variational Bayesian inference

UR - https://arxiv.org/abs/1605.04435

M3 - Conference article in proceedings

BT - Proceedings of the 5th Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI) at NIPS 2015

A2 - Rish, Irina

A2 - Wehbe, Leila

A2 - Langs, Georg

A2 - Grosse-Wentrup, Moritz

A2 - Murphy, Brian

A2 - Cecchi, Guillermo

CY - https://arxiv.org/abs/1605.04435

T2 - Workshop on Machine Learning and Interpretation in Neuroimaging

Y2 - 11 December 2015 through 12 December 2015

ER -

Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers

Abstract

Workshop

Keywords

Access to Document

Other files and links

Fingerprint

Science-IT

Cite this

Classification of weak multi-view signals by sharing factors in a mixture of Bayesian group factor analyzers

Abstract

Workshop

Keywords

Access to Document

Other files and links

Fingerprint

Equipment

Science-IT

Cite this