Scalable group level probabilistic sparse factor analysis

Jesper Løve Hinrich, Søren Føns Vind Nielsen, Nicolai Andre Brogaard Riis, Casper Eriksen, Jacob Frøsig, Marco D. F. Kristensen, Mikkel Nørgaard Schmidt, Kristoffer Hougaard Madsen, Morten Mørup

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Abstract

Many data-driven approaches exist to extract neural representations of functional magnetic resonance imaging (fMRI) data, but most of them lack a proper probabilistic formulation. We propose a scalable group level probabilistic sparse factor analysis (psFA) allowing spatially sparse maps, component pruning using automatic relevance determination (ARD) and subject specific heteroscedastic spatial noise modeling. For task-based and resting state fMRI, we show that the sparsity constraint gives rise to components similar to those obtained by group independent component analysis. The noise modeling shows that noise is reduced in areas typically associated with activation by the experimental design. The psFA model identifies sparse components and the probabilistic setting provides a natural way to handle parameter uncertainties. The variational Bayesian framework easily extends to more complex noise models than the presently considered.
Original languageEnglish
Title of host publicationProceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing
PublisherIEEE
Publication date2017
Pages6314-6318
ISBN (Print)9781509041169
DOIs
Publication statusPublished - 2017
Event42nd IEEE International Conference on Acoustics, Speech and Signal Processing: The internet of signals - Hilton New Orleans Riverside, New Orleans, United States
Duration: 5 Mar 20179 Mar 2017
Conference number: 42
http://www.ieee-icassp2017.org/

Conference

Conference42nd IEEE International Conference on Acoustics, Speech and Signal Processing
Number42
LocationHilton New Orleans Riverside
CountryUnited States
CityNew Orleans
Period05/03/201709/03/2017
Internet address
SeriesI E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings
ISSN1520-6149

Keywords

  • Probabilistic logic
  • Principal component analysis
  • Analytical models
  • Data models
  • Correlation
  • Indexes
  • Graphics processing units
  • sparsity
  • Probabilistic factor analysis
  • Neuroimaging
  • Variational Bayes
  • Independent components

Cite this

Hinrich, J. L., Nielsen, S. F. V., Riis, N. A. B., Eriksen, C., Frøsig, J., Kristensen, M. D. F., ... Mørup, M. (2017). Scalable group level probabilistic sparse factor analysis. In Proceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (pp. 6314-6318). IEEE. I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings https://doi.org/10.1109/ICASSP.2017.7953371
Hinrich, Jesper Løve ; Nielsen, Søren Føns Vind ; Riis, Nicolai Andre Brogaard ; Eriksen, Casper ; Frøsig, Jacob ; Kristensen, Marco D. F. ; Schmidt, Mikkel Nørgaard ; Madsen, Kristoffer Hougaard ; Mørup, Morten. / Scalable group level probabilistic sparse factor analysis. Proceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2017. pp. 6314-6318 (I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings).
@inproceedings{674932b9e9f349d3a33c0bb36bf1334c,
title = "Scalable group level probabilistic sparse factor analysis",
abstract = "Many data-driven approaches exist to extract neural representations of functional magnetic resonance imaging (fMRI) data, but most of them lack a proper probabilistic formulation. We propose a scalable group level probabilistic sparse factor analysis (psFA) allowing spatially sparse maps, component pruning using automatic relevance determination (ARD) and subject specific heteroscedastic spatial noise modeling. For task-based and resting state fMRI, we show that the sparsity constraint gives rise to components similar to those obtained by group independent component analysis. The noise modeling shows that noise is reduced in areas typically associated with activation by the experimental design. The psFA model identifies sparse components and the probabilistic setting provides a natural way to handle parameter uncertainties. The variational Bayesian framework easily extends to more complex noise models than the presently considered.",
keywords = "Probabilistic logic, Principal component analysis, Analytical models, Data models, Correlation, Indexes, Graphics processing units, sparsity, Probabilistic factor analysis, Neuroimaging, Variational Bayes, Independent components",
author = "Hinrich, {Jesper L{\o}ve} and Nielsen, {S{\o}ren F{\o}ns Vind} and Riis, {Nicolai Andre Brogaard} and Casper Eriksen and Jacob Fr{\o}sig and Kristensen, {Marco D. F.} and Schmidt, {Mikkel N{\o}rgaard} and Madsen, {Kristoffer Hougaard} and Morten M{\o}rup",
year = "2017",
doi = "10.1109/ICASSP.2017.7953371",
language = "English",
isbn = "9781509041169",
pages = "6314--6318",
booktitle = "Proceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing",
publisher = "IEEE",
address = "United States",

}

Hinrich, JL, Nielsen, SFV, Riis, NAB, Eriksen, C, Frøsig, J, Kristensen, MDF, Schmidt, MN, Madsen, KH & Mørup, M 2017, Scalable group level probabilistic sparse factor analysis. in Proceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings, pp. 6314-6318, 42nd IEEE International Conference on Acoustics, Speech and Signal Processing, New Orleans, United States, 05/03/2017. https://doi.org/10.1109/ICASSP.2017.7953371

Scalable group level probabilistic sparse factor analysis. / Hinrich, Jesper Løve; Nielsen, Søren Føns Vind; Riis, Nicolai Andre Brogaard; Eriksen, Casper; Frøsig, Jacob; Kristensen, Marco D. F.; Schmidt, Mikkel Nørgaard; Madsen, Kristoffer Hougaard; Mørup, Morten.

Proceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2017. p. 6314-6318 (I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings).

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

TY - GEN

T1 - Scalable group level probabilistic sparse factor analysis

AU - Hinrich, Jesper Løve

AU - Nielsen, Søren Føns Vind

AU - Riis, Nicolai Andre Brogaard

AU - Eriksen, Casper

AU - Frøsig, Jacob

AU - Kristensen, Marco D. F.

AU - Schmidt, Mikkel Nørgaard

AU - Madsen, Kristoffer Hougaard

AU - Mørup, Morten

PY - 2017

Y1 - 2017

N2 - Many data-driven approaches exist to extract neural representations of functional magnetic resonance imaging (fMRI) data, but most of them lack a proper probabilistic formulation. We propose a scalable group level probabilistic sparse factor analysis (psFA) allowing spatially sparse maps, component pruning using automatic relevance determination (ARD) and subject specific heteroscedastic spatial noise modeling. For task-based and resting state fMRI, we show that the sparsity constraint gives rise to components similar to those obtained by group independent component analysis. The noise modeling shows that noise is reduced in areas typically associated with activation by the experimental design. The psFA model identifies sparse components and the probabilistic setting provides a natural way to handle parameter uncertainties. The variational Bayesian framework easily extends to more complex noise models than the presently considered.

AB - Many data-driven approaches exist to extract neural representations of functional magnetic resonance imaging (fMRI) data, but most of them lack a proper probabilistic formulation. We propose a scalable group level probabilistic sparse factor analysis (psFA) allowing spatially sparse maps, component pruning using automatic relevance determination (ARD) and subject specific heteroscedastic spatial noise modeling. For task-based and resting state fMRI, we show that the sparsity constraint gives rise to components similar to those obtained by group independent component analysis. The noise modeling shows that noise is reduced in areas typically associated with activation by the experimental design. The psFA model identifies sparse components and the probabilistic setting provides a natural way to handle parameter uncertainties. The variational Bayesian framework easily extends to more complex noise models than the presently considered.

KW - Probabilistic logic

KW - Principal component analysis

KW - Analytical models

KW - Data models

KW - Correlation

KW - Indexes

KW - Graphics processing units

KW - sparsity

KW - Probabilistic factor analysis

KW - Neuroimaging

KW - Variational Bayes

KW - Independent components

U2 - 10.1109/ICASSP.2017.7953371

DO - 10.1109/ICASSP.2017.7953371

M3 - Article in proceedings

SN - 9781509041169

SP - 6314

EP - 6318

BT - Proceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing

PB - IEEE

ER -

Hinrich JL, Nielsen SFV, Riis NAB, Eriksen C, Frøsig J, Kristensen MDF et al. Scalable group level probabilistic sparse factor analysis. In Proceedings of 2017 IEEE International Conference on Acoustics, Speech and Signal Processing. IEEE. 2017. p. 6314-6318. (I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings). https://doi.org/10.1109/ICASSP.2017.7953371