GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model

Yanyan Sheng, William S. Welling, Michelle Zhu

Research output: Chapter in Book/Report/Conference proceedingChapterResearchpeer-review

Abstract

Item response theory (IRT) is a popular approach used for addressing large-scale statistical problems in psychometrics as well as in other fields. The fully Bayesian approach for estimating IRT models is usually memory and computational expensive due to the large number of iterations. This limits the use of the procedure in many applications. In an effort to overcome such restrictions, previous studies proposed to tackle the problem using massive core-based graphic processing units (GPU), and demonstrated the advantage of this approach over the message passing interface (MPI) by showing that a single GPU card could achieve a speedup of up to 50×. Given that GPU is practical, cost-effective, and convenient, this study aims to seek further improvements using a single GPU card.

Original languageEnglish
Title of host publicationQuantitative Psychology Research
Subtitle of host publicationThe 79th Annual Meeting of the Psychometric Society, Madison, Wisconsin, 2014
PublisherSpringer International Publishing
Pages59-73
Number of pages15
Volume140
ISBN (Electronic)9783319199771
ISBN (Print)9783319199764
DOIs
StatePublished - 8 Aug 2015

Fingerprint

Bayes Theorem
model theory
Psychometrics
Costs and Cost Analysis
psychometrics
costs

Keywords

  • Bayesian estimation
  • CUDA
  • High performance computing
  • Item response theory
  • MCMC
  • Optimization
  • Two-parameter IRT model

Cite this

Sheng, Y., Welling, W. S., & Zhu, M. (2015). GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model. In Quantitative Psychology Research: The 79th Annual Meeting of the Psychometric Society, Madison, Wisconsin, 2014 (Vol. 140, pp. 59-73). Springer International Publishing. https://doi.org/10.1007/978-3-319-19977-1_5
Sheng, Yanyan ; Welling, William S. ; Zhu, Michelle. / GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model. Quantitative Psychology Research: The 79th Annual Meeting of the Psychometric Society, Madison, Wisconsin, 2014. Vol. 140 Springer International Publishing, 2015. pp. 59-73
@inbook{60bbc7ba31424468a37c93332f723c4c,
title = "GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model",
abstract = "Item response theory (IRT) is a popular approach used for addressing large-scale statistical problems in psychometrics as well as in other fields. The fully Bayesian approach for estimating IRT models is usually memory and computational expensive due to the large number of iterations. This limits the use of the procedure in many applications. In an effort to overcome such restrictions, previous studies proposed to tackle the problem using massive core-based graphic processing units (GPU), and demonstrated the advantage of this approach over the message passing interface (MPI) by showing that a single GPU card could achieve a speedup of up to 50×. Given that GPU is practical, cost-effective, and convenient, this study aims to seek further improvements using a single GPU card.",
keywords = "Bayesian estimation, CUDA, High performance computing, Item response theory, MCMC, Optimization, Two-parameter IRT model",
author = "Yanyan Sheng and Welling, {William S.} and Michelle Zhu",
year = "2015",
month = "8",
day = "8",
doi = "10.1007/978-3-319-19977-1_5",
language = "English",
isbn = "9783319199764",
volume = "140",
pages = "59--73",
booktitle = "Quantitative Psychology Research",
publisher = "Springer International Publishing",

}

Sheng, Y, Welling, WS & Zhu, M 2015, GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model. in Quantitative Psychology Research: The 79th Annual Meeting of the Psychometric Society, Madison, Wisconsin, 2014. vol. 140, Springer International Publishing, pp. 59-73. https://doi.org/10.1007/978-3-319-19977-1_5

GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model. / Sheng, Yanyan; Welling, William S.; Zhu, Michelle.

Quantitative Psychology Research: The 79th Annual Meeting of the Psychometric Society, Madison, Wisconsin, 2014. Vol. 140 Springer International Publishing, 2015. p. 59-73.

Research output: Chapter in Book/Report/Conference proceedingChapterResearchpeer-review

TY - CHAP

T1 - GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model

AU - Sheng, Yanyan

AU - Welling, William S.

AU - Zhu, Michelle

PY - 2015/8/8

Y1 - 2015/8/8

N2 - Item response theory (IRT) is a popular approach used for addressing large-scale statistical problems in psychometrics as well as in other fields. The fully Bayesian approach for estimating IRT models is usually memory and computational expensive due to the large number of iterations. This limits the use of the procedure in many applications. In an effort to overcome such restrictions, previous studies proposed to tackle the problem using massive core-based graphic processing units (GPU), and demonstrated the advantage of this approach over the message passing interface (MPI) by showing that a single GPU card could achieve a speedup of up to 50×. Given that GPU is practical, cost-effective, and convenient, this study aims to seek further improvements using a single GPU card.

AB - Item response theory (IRT) is a popular approach used for addressing large-scale statistical problems in psychometrics as well as in other fields. The fully Bayesian approach for estimating IRT models is usually memory and computational expensive due to the large number of iterations. This limits the use of the procedure in many applications. In an effort to overcome such restrictions, previous studies proposed to tackle the problem using massive core-based graphic processing units (GPU), and demonstrated the advantage of this approach over the message passing interface (MPI) by showing that a single GPU card could achieve a speedup of up to 50×. Given that GPU is practical, cost-effective, and convenient, this study aims to seek further improvements using a single GPU card.

KW - Bayesian estimation

KW - CUDA

KW - High performance computing

KW - Item response theory

KW - MCMC

KW - Optimization

KW - Two-parameter IRT model

UR - http://www.scopus.com/inward/record.url?scp=84955387097&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-19977-1_5

DO - 10.1007/978-3-319-19977-1_5

M3 - Chapter

SN - 9783319199764

VL - 140

SP - 59

EP - 73

BT - Quantitative Psychology Research

PB - Springer International Publishing

ER -

Sheng Y, Welling WS, Zhu M. GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model. In Quantitative Psychology Research: The 79th Annual Meeting of the Psychometric Society, Madison, Wisconsin, 2014. Vol. 140. Springer International Publishing. 2015. p. 59-73 https://doi.org/10.1007/978-3-319-19977-1_5