Accelerating collaborative filtering using concepts from high performance computing

Mark Gates; Hartwig Anzt; Jakub Kurzak; Jack Dongarra

doi:10.1109/BigData.2015.7363811

Accelerating collaborative filtering using concepts from high performance computing

Mark Gates, Hartwig Anzt, Jakub Kurzak, Jack Dongarra

University of Tennessee, Knoxville

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

28 Scopus citations

Abstract

In this paper we accelerate the Alternating Least Squares (ALS) algorithm used for generating product recommendations on the basis of implicit feedback datasets. We approach the algorithm with concepts proven to be successful in High Performance Computing. This includes the formulation of the algorithm as a mix of cache-optimized algorithm-specific kernels and standard BLAS routines, acceleration via graphics processing units (GPUs), use of parallel batched kernels, and autotuning to identify performance winners. For benchmark datasets, the multi-threaded CPU implementation we propose achieves more than a 10 times speedup over the implementations available in the GraphLab and Spark MLlib software packages. For the GPU implementation, the parameters of an algorithm-specific kernel were optimized using a comprehensive autotuning sweep. This results in an additional 2 times speedup over our CPU implementation.

Original language	English
Title of host publication	Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015
Editors	Feng Luo, Kemafor Ogan, Mohammed J. Zaki, Laura Haas, Beng Chin Ooi, Vipin Kumar, Sudarsan Rachuri, Saumyadipta Pyne, Howard Ho, Xiaohua Hu, Shipeng Yu, Morris Hui-I Hsiao, Jian Li
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	667-676
Number of pages	10
ISBN (Electronic)	9781479999255
DOIs	https://doi.org/10.1109/BigData.2015.7363811
State	Published - Dec 22 2015
Externally published	Yes
Event	3rd IEEE International Conference on Big Data, IEEE Big Data 2015 - Santa Clara, United States Duration: Oct 29 2015 → Nov 1 2015

Publication series

Name	Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015

Conference

Conference	3rd IEEE International Conference on Big Data, IEEE Big Data 2015
Country/Territory	United States
City	Santa Clara
Period	10/29/15 → 11/1/15

Bibliographical note

Publisher Copyright:
© 2015 IEEE.

Keywords

Alternating Least Squares
Autotuning
Batched Cholesky
Collaborative Filtering
GPUs

Access to Document

10.1109/BigData.2015.7363811

Cite this

Gates, M., Anzt, H., Kurzak, J., & Dongarra, J. (2015). Accelerating collaborative filtering using concepts from high performance computing. In F. Luo, K. Ogan, M. J. Zaki, L. Haas, B. C. Ooi, V. Kumar, S. Rachuri, S. Pyne, H. Ho, X. Hu, S. Yu, M. H.-I. Hsiao, & J. Li (Eds.), Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015 (pp. 667-676). Article 7363811 (Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/BigData.2015.7363811

Gates, Mark ; Anzt, Hartwig ; Kurzak, Jakub et al. / Accelerating collaborative filtering using concepts from high performance computing. Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015. editor / Feng Luo ; Kemafor Ogan ; Mohammed J. Zaki ; Laura Haas ; Beng Chin Ooi ; Vipin Kumar ; Sudarsan Rachuri ; Saumyadipta Pyne ; Howard Ho ; Xiaohua Hu ; Shipeng Yu ; Morris Hui-I Hsiao ; Jian Li. Institute of Electrical and Electronics Engineers Inc., 2015. pp. 667-676 (Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015).

@inproceedings{b57b4371e1d2474e9889f85c77dc3ab3,

title = "Accelerating collaborative filtering using concepts from high performance computing",

abstract = "In this paper we accelerate the Alternating Least Squares (ALS) algorithm used for generating product recommendations on the basis of implicit feedback datasets. We approach the algorithm with concepts proven to be successful in High Performance Computing. This includes the formulation of the algorithm as a mix of cache-optimized algorithm-specific kernels and standard BLAS routines, acceleration via graphics processing units (GPUs), use of parallel batched kernels, and autotuning to identify performance winners. For benchmark datasets, the multi-threaded CPU implementation we propose achieves more than a 10 times speedup over the implementations available in the GraphLab and Spark MLlib software packages. For the GPU implementation, the parameters of an algorithm-specific kernel were optimized using a comprehensive autotuning sweep. This results in an additional 2 times speedup over our CPU implementation.",

keywords = "Alternating Least Squares, Autotuning, Batched Cholesky, Collaborative Filtering, GPUs",

author = "Mark Gates and Hartwig Anzt and Jakub Kurzak and Jack Dongarra",

note = "Publisher Copyright: {\textcopyright} 2015 IEEE.; 3rd IEEE International Conference on Big Data, IEEE Big Data 2015 ; Conference date: 29-10-2015 Through 01-11-2015",

year = "2015",

month = dec,

day = "22",

doi = "10.1109/BigData.2015.7363811",

language = "English",

series = "Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "667--676",

editor = "Feng Luo and Kemafor Ogan and Zaki, {Mohammed J.} and Laura Haas and Ooi, {Beng Chin} and Vipin Kumar and Sudarsan Rachuri and Saumyadipta Pyne and Howard Ho and Xiaohua Hu and Shipeng Yu and Hsiao, {Morris Hui-I} and Jian Li",

booktitle = "Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015",

}

Gates, M, Anzt, H, Kurzak, J & Dongarra, J 2015, Accelerating collaborative filtering using concepts from high performance computing. in F Luo, K Ogan, MJ Zaki, L Haas, BC Ooi, V Kumar, S Rachuri, S Pyne, H Ho, X Hu, S Yu, MH-I Hsiao & J Li (eds), Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015., 7363811, Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015, Institute of Electrical and Electronics Engineers Inc., pp. 667-676, 3rd IEEE International Conference on Big Data, IEEE Big Data 2015, Santa Clara, United States, 10/29/15. https://doi.org/10.1109/BigData.2015.7363811

Accelerating collaborative filtering using concepts from high performance computing. / Gates, Mark; Anzt, Hartwig; Kurzak, Jakub et al.
Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015. ed. / Feng Luo; Kemafor Ogan; Mohammed J. Zaki; Laura Haas; Beng Chin Ooi; Vipin Kumar; Sudarsan Rachuri; Saumyadipta Pyne; Howard Ho; Xiaohua Hu; Shipeng Yu; Morris Hui-I Hsiao; Jian Li. Institute of Electrical and Electronics Engineers Inc., 2015. p. 667-676 7363811 (Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015).

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

TY - GEN

T1 - Accelerating collaborative filtering using concepts from high performance computing

AU - Gates, Mark

AU - Anzt, Hartwig

AU - Kurzak, Jakub

AU - Dongarra, Jack

PY - 2015/12/22

Y1 - 2015/12/22

N2 - In this paper we accelerate the Alternating Least Squares (ALS) algorithm used for generating product recommendations on the basis of implicit feedback datasets. We approach the algorithm with concepts proven to be successful in High Performance Computing. This includes the formulation of the algorithm as a mix of cache-optimized algorithm-specific kernels and standard BLAS routines, acceleration via graphics processing units (GPUs), use of parallel batched kernels, and autotuning to identify performance winners. For benchmark datasets, the multi-threaded CPU implementation we propose achieves more than a 10 times speedup over the implementations available in the GraphLab and Spark MLlib software packages. For the GPU implementation, the parameters of an algorithm-specific kernel were optimized using a comprehensive autotuning sweep. This results in an additional 2 times speedup over our CPU implementation.

AB - In this paper we accelerate the Alternating Least Squares (ALS) algorithm used for generating product recommendations on the basis of implicit feedback datasets. We approach the algorithm with concepts proven to be successful in High Performance Computing. This includes the formulation of the algorithm as a mix of cache-optimized algorithm-specific kernels and standard BLAS routines, acceleration via graphics processing units (GPUs), use of parallel batched kernels, and autotuning to identify performance winners. For benchmark datasets, the multi-threaded CPU implementation we propose achieves more than a 10 times speedup over the implementations available in the GraphLab and Spark MLlib software packages. For the GPU implementation, the parameters of an algorithm-specific kernel were optimized using a comprehensive autotuning sweep. This results in an additional 2 times speedup over our CPU implementation.

KW - Alternating Least Squares

KW - Autotuning

KW - Batched Cholesky

KW - Collaborative Filtering

KW - GPUs

UR - http://www.scopus.com/inward/record.url?scp=84963785643&partnerID=8YFLogxK

U2 - 10.1109/BigData.2015.7363811

DO - 10.1109/BigData.2015.7363811

M3 - Conference contribution

AN - SCOPUS:84963785643

T3 - Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015

SP - 667

EP - 676

BT - Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015

A2 - Luo, Feng

A2 - Ogan, Kemafor

A2 - Zaki, Mohammed J.

A2 - Haas, Laura

A2 - Ooi, Beng Chin

A2 - Kumar, Vipin

A2 - Rachuri, Sudarsan

A2 - Pyne, Saumyadipta

A2 - Ho, Howard

A2 - Hu, Xiaohua

A2 - Yu, Shipeng

A2 - Hsiao, Morris Hui-I

A2 - Li, Jian

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 3rd IEEE International Conference on Big Data, IEEE Big Data 2015

Y2 - 29 October 2015 through 1 November 2015

ER -

Gates M, Anzt H, Kurzak J, Dongarra J. Accelerating collaborative filtering using concepts from high performance computing. In Luo F, Ogan K, Zaki MJ, Haas L, Ooi BC, Kumar V, Rachuri S, Pyne S, Ho H, Hu X, Yu S, Hsiao MHI, Li J, editors, Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015. Institute of Electrical and Electronics Engineers Inc. 2015. p. 667-676. 7363811. (Proceedings - 2015 IEEE International Conference on Big Data, IEEE Big Data 2015). doi: 10.1109/BigData.2015.7363811

Accelerating collaborative filtering using concepts from high performance computing

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this