Turbo-SMT: Accelerating coupled sparse matrix-tensor factorizations by 200x

Evangelos E. Papalexakis; Christos Faloutsos; Tom M. Mitchell; Partha Pratim Talukdar; Nicholas D. Sidiropoulos; Brian Murphy

doi:10.1137/1.9781611973440.14

Turbo-SMT: Accelerating coupled sparse matrix-tensor factorizations by 200x

Evangelos E. Papalexakis, Christos Faloutsos, Tom M. Mitchell, Partha Pratim Talukdar, Nicholas D. Sidiropoulos, Brian Murphy

Electrical and Computer Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

33 Scopus citations

Abstract

How can we correlate the neural activity in the human brain as it responds to typed words, with properties of these terms (like 'edible', 'fits in hand')? In short, we want to find latent variables, that jointly explain both the brain activity, as well as the behavioral responses. This is one of many settings of the Coupled Matrix-Tensor Factorization (CMTF) problem. Can we accelerate any CMTF solver, so that it runs within a few minutes instead of tens of hours to a day, while maintaining good accuracy? We introduce TURBO-SMT, a meta-method capable of doing exactly that: it boosts the performance of any CMTF algorithm, by up to 200x, along with an up to 65 fold increase in sparsity, with comparable accuracy to the baseline. We apply TURBO-SMT to BRAIN Q, a dataset consisting of a (nouns, brain voxels, human subjects) tensor and a (nouns, properties) matrix, with coupling along the nouns dimension. TURBO-SMT is able to find meaningful latent variables, as well as to predict brain activity with competitive accuracy.

Original language	English (US)
Title of host publication	SIAM International Conference on Data Mining 2014, SDM 2014
Editors	Mohammed J. Zaki, Arindam Banerjee, Srinivasan Parthasarathy, Pang Ning-Tan, Zoran Obradovic, Chandrika Kamath
Publisher	Society for Industrial and Applied Mathematics Publications
Pages	118-126
Number of pages	9
ISBN (Electronic)	9781510811515
DOIs	https://doi.org/10.1137/1.9781611973440.14
State	Published - 2014
Event	14th SIAM International Conference on Data Mining, SDM 2014 - Philadelphia, United States Duration: Apr 24 2014 → Apr 26 2014

Publication series

Name	SIAM International Conference on Data Mining 2014, SDM 2014
Volume	1

Other

Other	14th SIAM International Conference on Data Mining, SDM 2014
Country/Territory	United States
City	Philadelphia
Period	4/24/14 → 4/26/14

Bibliographical note

Funding Information:
Research was funded by grants NSF IIS-1247489, NSF IIS- 1247632, NSF CDI 0835797, NIH/NICHD 12165321, and DARPA FA87501320005. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the funding parties. The authors would also like to thank Leila Wehbe and Alona Fyshe for their initial help with the BrainQ data.

Publisher Copyright:
Copyright © SIAM.

Publisher link

10.1137/1.9781611973440.14

Find It

Find It at U of M Libraries

Cite this

Papalexakis, E. E., Faloutsos, C., Mitchell, T. M., Talukdar, P. P., Sidiropoulos, N. D., & Murphy, B. (2014). Turbo-SMT: Accelerating coupled sparse matrix-tensor factorizations by 200x. In M. J. Zaki, A. Banerjee, S. Parthasarathy, P. Ning-Tan, Z. Obradovic, & C. Kamath (Eds.), SIAM International Conference on Data Mining 2014, SDM 2014 (pp. 118-126). (SIAM International Conference on Data Mining 2014, SDM 2014; Vol. 1). Society for Industrial and Applied Mathematics Publications. https://doi.org/10.1137/1.9781611973440.14

Turbo-SMT : Accelerating coupled sparse matrix-tensor factorizations by 200x. / Papalexakis, Evangelos E.; Faloutsos, Christos; Mitchell, Tom M. et al.

SIAM International Conference on Data Mining 2014, SDM 2014. ed. / Mohammed J. Zaki; Arindam Banerjee; Srinivasan Parthasarathy; Pang Ning-Tan; Zoran Obradovic; Chandrika Kamath. Society for Industrial and Applied Mathematics Publications, 2014. p. 118-126 (SIAM International Conference on Data Mining 2014, SDM 2014; Vol. 1).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Papalexakis, EE, Faloutsos, C, Mitchell, TM, Talukdar, PP, Sidiropoulos, ND & Murphy, B 2014, Turbo-SMT: Accelerating coupled sparse matrix-tensor factorizations by 200x. in MJ Zaki, A Banerjee, S Parthasarathy, P Ning-Tan, Z Obradovic & C Kamath (eds), SIAM International Conference on Data Mining 2014, SDM 2014. SIAM International Conference on Data Mining 2014, SDM 2014, vol. 1, Society for Industrial and Applied Mathematics Publications, pp. 118-126, 14th SIAM International Conference on Data Mining, SDM 2014, Philadelphia, United States, 4/24/14. https://doi.org/10.1137/1.9781611973440.14

Papalexakis EE, Faloutsos C, Mitchell TM, Talukdar PP, Sidiropoulos ND, Murphy B. Turbo-SMT: Accelerating coupled sparse matrix-tensor factorizations by 200x. In Zaki MJ, Banerjee A, Parthasarathy S, Ning-Tan P, Obradovic Z, Kamath C, editors, SIAM International Conference on Data Mining 2014, SDM 2014. Society for Industrial and Applied Mathematics Publications. 2014. p. 118-126. (SIAM International Conference on Data Mining 2014, SDM 2014). doi: 10.1137/1.9781611973440.14

Papalexakis, Evangelos E. ; Faloutsos, Christos ; Mitchell, Tom M. et al. / Turbo-SMT : Accelerating coupled sparse matrix-tensor factorizations by 200x. SIAM International Conference on Data Mining 2014, SDM 2014. editor / Mohammed J. Zaki ; Arindam Banerjee ; Srinivasan Parthasarathy ; Pang Ning-Tan ; Zoran Obradovic ; Chandrika Kamath. Society for Industrial and Applied Mathematics Publications, 2014. pp. 118-126 (SIAM International Conference on Data Mining 2014, SDM 2014).

@inproceedings{3466a0ad922d432285e7bcebf9b0757d,

title = "Turbo-SMT: Accelerating coupled sparse matrix-tensor factorizations by 200x",

abstract = "How can we correlate the neural activity in the human brain as it responds to typed words, with properties of these terms (like 'edible', 'fits in hand')? In short, we want to find latent variables, that jointly explain both the brain activity, as well as the behavioral responses. This is one of many settings of the Coupled Matrix-Tensor Factorization (CMTF) problem. Can we accelerate any CMTF solver, so that it runs within a few minutes instead of tens of hours to a day, while maintaining good accuracy? We introduce TURBO-SMT, a meta-method capable of doing exactly that: it boosts the performance of any CMTF algorithm, by up to 200x, along with an up to 65 fold increase in sparsity, with comparable accuracy to the baseline. We apply TURBO-SMT to BRAIN Q, a dataset consisting of a (nouns, brain voxels, human subjects) tensor and a (nouns, properties) matrix, with coupling along the nouns dimension. TURBO-SMT is able to find meaningful latent variables, as well as to predict brain activity with competitive accuracy.",

author = "Papalexakis, {Evangelos E.} and Christos Faloutsos and Mitchell, {Tom M.} and Talukdar, {Partha Pratim} and Sidiropoulos, {Nicholas D.} and Brian Murphy",

note = "Funding Information: Research was funded by grants NSF IIS-1247489, NSF IIS- 1247632, NSF CDI 0835797, NIH/NICHD 12165321, and DARPA FA87501320005. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the funding parties. The authors would also like to thank Leila Wehbe and Alona Fyshe for their initial help with the BrainQ data. Publisher Copyright: Copyright {\textcopyright} SIAM.; 14th SIAM International Conference on Data Mining, SDM 2014 ; Conference date: 24-04-2014 Through 26-04-2014",

year = "2014",

doi = "10.1137/1.9781611973440.14",

language = "English (US)",

series = "SIAM International Conference on Data Mining 2014, SDM 2014",

publisher = "Society for Industrial and Applied Mathematics Publications",

pages = "118--126",

editor = "Zaki, {Mohammed J.} and Arindam Banerjee and Srinivasan Parthasarathy and Pang Ning-Tan and Zoran Obradovic and Chandrika Kamath",

booktitle = "SIAM International Conference on Data Mining 2014, SDM 2014",

}

TY - GEN

T1 - Turbo-SMT

T2 - 14th SIAM International Conference on Data Mining, SDM 2014

AU - Papalexakis, Evangelos E.

AU - Faloutsos, Christos

AU - Mitchell, Tom M.

AU - Talukdar, Partha Pratim

AU - Sidiropoulos, Nicholas D.

AU - Murphy, Brian

N1 - Funding Information: Research was funded by grants NSF IIS-1247489, NSF IIS- 1247632, NSF CDI 0835797, NIH/NICHD 12165321, and DARPA FA87501320005. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the funding parties. The authors would also like to thank Leila Wehbe and Alona Fyshe for their initial help with the BrainQ data. Publisher Copyright: Copyright © SIAM.

PY - 2014

Y1 - 2014

N2 - How can we correlate the neural activity in the human brain as it responds to typed words, with properties of these terms (like 'edible', 'fits in hand')? In short, we want to find latent variables, that jointly explain both the brain activity, as well as the behavioral responses. This is one of many settings of the Coupled Matrix-Tensor Factorization (CMTF) problem. Can we accelerate any CMTF solver, so that it runs within a few minutes instead of tens of hours to a day, while maintaining good accuracy? We introduce TURBO-SMT, a meta-method capable of doing exactly that: it boosts the performance of any CMTF algorithm, by up to 200x, along with an up to 65 fold increase in sparsity, with comparable accuracy to the baseline. We apply TURBO-SMT to BRAIN Q, a dataset consisting of a (nouns, brain voxels, human subjects) tensor and a (nouns, properties) matrix, with coupling along the nouns dimension. TURBO-SMT is able to find meaningful latent variables, as well as to predict brain activity with competitive accuracy.

AB - How can we correlate the neural activity in the human brain as it responds to typed words, with properties of these terms (like 'edible', 'fits in hand')? In short, we want to find latent variables, that jointly explain both the brain activity, as well as the behavioral responses. This is one of many settings of the Coupled Matrix-Tensor Factorization (CMTF) problem. Can we accelerate any CMTF solver, so that it runs within a few minutes instead of tens of hours to a day, while maintaining good accuracy? We introduce TURBO-SMT, a meta-method capable of doing exactly that: it boosts the performance of any CMTF algorithm, by up to 200x, along with an up to 65 fold increase in sparsity, with comparable accuracy to the baseline. We apply TURBO-SMT to BRAIN Q, a dataset consisting of a (nouns, brain voxels, human subjects) tensor and a (nouns, properties) matrix, with coupling along the nouns dimension. TURBO-SMT is able to find meaningful latent variables, as well as to predict brain activity with competitive accuracy.

UR - http://www.scopus.com/inward/record.url?scp=84959875939&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84959875939&partnerID=8YFLogxK

U2 - 10.1137/1.9781611973440.14

DO - 10.1137/1.9781611973440.14

M3 - Conference contribution

AN - SCOPUS:84959875939

T3 - SIAM International Conference on Data Mining 2014, SDM 2014

SP - 118

EP - 126

BT - SIAM International Conference on Data Mining 2014, SDM 2014

A2 - Zaki, Mohammed J.

A2 - Banerjee, Arindam

A2 - Parthasarathy, Srinivasan

A2 - Ning-Tan, Pang

A2 - Obradovic, Zoran

A2 - Kamath, Chandrika

PB - Society for Industrial and Applied Mathematics Publications

Y2 - 24 April 2014 through 26 April 2014

ER -

Turbo-SMT: Accelerating coupled sparse matrix-tensor factorizations by 200x

Abstract

Publication series

Other

Bibliographical note

Publisher link

Find It

Other files and links

Fingerprint

Cite this