Abstract
Item response theory (IRT) is a popular approach for addressing large-scale assessment problems in psychometrics and other areas of applied research. An emergent research direction that integrates it with machine learning techniques has made IRT applicable to a wide range of fields. The fully Bayesian approach for estimating IRT models is computationally expensive due to the large number of iterations, which require a large amount of memory to store massive amount of data. This limits the use of the procedure in many applications using traditional CPU architecture. In an effort to overcome such restrictions, previous studies focused on utilizing high performance computing using either distributed memory-based Message Passing Interface (MPI) or massive threads compute unified device architecture (CUDA) to achieve certain speedups with a simple IRT model. This study focuses on this model and aims at demonstrating the scalability of parallel algorithms integrating CUDA into MPI computing paradigm.
Similar content being viewed by others
Data availability
Enquiries about data availability should be directed to the authors.
References
Bock, R.D., Aitkin, M.: Marginal maximum likelihood estimation of item parameters: application of an EM algorithm. Psychometrika 46(4), 443–459 (1981)
Mislevy, R.J.: Estimation of latent group effects. J. Am. Stat. Assoc. 80(392), 993–997 (1985)
Patz, R.J., Junker, B.: A straightforward approach to Markov chain Monte Carlo methods for item response model. J. Educ. Behav. Stat. 24(2), 146–178 (1999)
Tutakawa, R.K., Lin, H.Y.: Bayesian estimation of item response curves. Psychometrika 51(2), 251–267 (1986)
Bafumi, J., Gelman, A., Park, D.K., Kaplan, N.: Practical issues in implementing and understanding Bayesian ideal point estimation. Polit. Anal. 13(2), 171–187 (2005)
Martin, C.S., Chung, T., Kirisci, L., Langenbucher, J.W.: Item response theory analysis of diagnostic criteria for alcohol and cannabis use disorders in adolescents: implications for DSM-V. J. Abnorm. Psychol. 115(4), 807–814 (2006)
Feske, U., Kirisci, L., Tarter, R.E., Plkonis, P.A.: An application of item response theory to the DSM-III-R criteria for borderline personality disorder. J. Pers. Disord. 21(4), 418–433 (2007)
Beseler, C.L., Taylor, L.A., Leeman, R.F.: An item-response theory analysis of DSM-IV alcohol-use disorder criteria and “binge’’ drinking in undergraduates. J. Stud. Alcohol Drugs 71(3), 418–423 (2010)
Gilder, D.A., Gizer, I.R., Ehlers, C.L.: Item response theory analysis of binge drinking and its relationship to lifetime alcohol use disorder symptom severity in an American Indian community sample. Alcoholism 35(5), 984–995 (2011)
Panter, A.T., Reeve, B.B.: Assessing tobacco beliefs among youth using item response theory models. Drug Alcohol Depend. 68(1), 21–39 (2002)
Courvoisier, D., Etter, J.F.: Using item response theory to study the convergent and discriminant validity of three questionnaires measuring cigarette dependence. Psychol. Addict. Behav. 22(3), 391–401 (2008)
Rose, J.S., Dierker, L.C.: An item response theory analysis of nicotine dependence symptoms in recent onset adolescent smokers. Drug Alcohol Depend. 110(12), 70–79 (2010)
Fienberg, S.E., Johnson, M.S., Junker, B.W.: Classical multilevel and Bayesian approaches to population size estimation using multiple lists. J. R. Stat. Soc. Ser. A 162(3), 383–392 (1999)
Reiser, M.: An application of the item-response model to psychiatric epidemiology. Sociol. Methods Res. 18(1), 66–103 (1989)
Orlando, M., Sherbourne, C.D., Thissen, D.: Summed-score linking using item response theory: application to depression measurement. Psychol. Assess. 12(3), 354–359 (2000)
Tsutsumi, A., Iwata, N., Watanabe, N., de Jonge, J., Pikhart, H., Fernndez-Lpez, J.A., Xu, L., Peter, R., Knutsson, A., Niedhammer, I., Kawakami, N., Siegrist, J.: Application of item response theory to achieve cross-cultural comparability of occupational stress measurement. Int. J. Methods Psychiatr. Res. 18(1), 58–67 (2009)
Chen, Z., Ahn, H.: Item response theory based ensemble in machine learning. Int. J. Autom. Comput. 17(5), 621–636 (2020)
Martínez-Plumed, F., Pruden̂cio, R. B. C., Martínez-Usó, A., Hernańdez-Orallo, J.: Making sense of item response theory in machine learning. In: ECAI’16: Proceedings of the Twenty-second European Conference on Artificial Intelligence, pp. 1140–1148 (2016)
Bergner, Y., Dröschler, S., Kortemeyer, G., Rayyan, S., Seaton, D., Pritchard, D. E.: Model-based collaborative filtering analysis of student response data: machine-learning item response theory. Paper presented at the International Conference on Educational Data Mining (EDM), Chania, Greece, June (2012)
Birnbaum, A.: Statistical theory for logistic mental test models with a prior distribution of ability. J. Math. Psychol. 6(2), 258–276 (1969)
Baker, F.B., Kim, S.H.: Item Response Theory: Parameter Estimation Techniques, 2nd edn. Dekker, New York (2004)
Molenaar, I.W.: Estimation of item parameters. In: Fischer, G.H., Molenaar, I.W. (eds.) Rasch Models: Foundations, Recent Developments, and Applications, pp. 39–51. Springer, New York (1995)
Smith, A.F.M., Roberts, G.O.: Bayesian computation via the Gibbs sampler and related Markov chain Monte Carlo methods (with discussion). J. R. Stat. Soc. B 55(1), 3–23 (1993)
Tierney, L.: Markov chains for exploring posterior distributions. Ann. Stat. 22(4), 1701–1728 (1994)
Albert, J.H.: Bayesian estimation of normal ogive item response curves using Gibbs sampling. J. Educ. Stat. 17(3), 251–269 (1992)
Geman, S., Geman, D.: Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images. IEEE Trans. Pattern Anal. Mach. Intell. 6(6), 721–741 (1984)
Lord, F.M., Novick, M.R.: Statistical Theories of Mental Test Scores. Addison-Wesley, Boston (1968)
Sheng, Y., Headrick, T.C.: An algorithm for implementing Gibbs sampling for 2PNO IRT models. J. Mod. Appl. Stat. Methods 6(1), 341–349 (2007)
Harwell, M., Stone, C.A., Hsu, H., Kirisci, L.: Monte Carlo studies in item response theory. Appl. Psychol. Meas. 20(2), 101–126 (1996)
Foster, I.: Designing and Building Parallel Programs: Concepts and Tools for Parallel Software Engineering. Addison-Wesley, Boston (1995)
Pastias, K., Rahimi, M., Sheng, Y., Rahimi, S.: Parallel computing with a Bayesian item response model. Am. J. Comput. Math. 2(2), 65–71 (2012)
Sheng, Y., Rahimi, M.: High performance Gibbs sampling for IRT models using row-wise decomposition. ISRN Comput. Math. 2012(264040), 1–9 (2012)
Sheng, Y., Welling, W.S., Zhu, M.M.: A GPU-based Gibbs sampler for a unidimensional IRT model. ISRN Comput. Math. 2014(368149), 1–11 (2014)
Sheng, Y., Welling, W.S., Zhu, M.M.: GPU-accelerated computing with Gibbs sampler for the 2PNO IRT model. In: van der Ark, L.A., et al. (eds.) Quantitative Psychology Research, pp. 59–73. Springer, New York (2015)
NVIDIA, Vingelmann, P., Fitzek, F.H.P.: CUDA, Release: 10.2.89 (2020) https://developer.nvidia.com/cuda-toolkit
Kraus, J.: An introduction to CUDA-aware MPI. NVidia Developer (blog) (2013) https://developer.nvidia.com/blog/introduction-cuda-aware-mpi/
Adinets, A.: CUDA dynamic parallelism API and principles. NVidia Developer (blog) (2018) https://developer.nvidia.com/blog/using-cuda-warp-level-primitives/
CUDA-CUBLAS Library 11.7. NVIDIA Corp. (2022) https://docs.nvidia.com/cuda/cublas/
Reckase, M.D.: The past and future of multidimensional item response theory. Appl. Psychol. Meas. 21(1), 25–36 (1997)
Farahani, H. S., Fatehi, A., Sh, M. A.: Between-domain instance transition via the process of Gibbs sampling in RBM. arXiv, (2020). https://doi.org/10.48550/arxiv.2006.14538
Papamarkou, T., Hinkle, J., Young, M. T., Womble, D.: Challenges in Markov chain Monte Carlo for Bayesian neural networks. arXiv, (2019). https://doi.org/10.48550/arxiv.1910.06539
Canillas, R., Laurent, R., Faix, M., Vaufreydaz, D., Mazer, E.: Autonomous robot controller using bitwise Gibbs sampling. In: 2016 IEEE 15th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC), pp. 72–76 (2016). https://doi.org/10.1109/ICCI-CC.2016.7862096
Acknowledgements
We thank the University of Chicago’s Research Computing Center for their support of this work.
Funding
No funding was received for conducting this study.
Author information
Authors and Affiliations
Contributions
WW and YS designed the study. WW developed and optimized the parallel programs and carried out the analyses. YS drafted the main manuscript text and prepared all figures. MZ reviewed the initial design and the developed parallel algorithms. All authors reviewed the manuscript.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix 1
Appendix 1
Figure 9 displays the execution time of CUDA-aware MPI with varying number of nodes and tasks to implement Gibbs sampling to tests with k (\(k=20, 50, 100, 200\)) items and 500 to 1,000,000 persons with different sizes, whereas Fig. 10 displays the speedup of CUDA-aware MPI over MPI under the same sample size, test length and node/task conditions. It is noted that MPI implementations with 1 node and 2 tasks failed to provide a result when sample size reaches 1,000,000, and hence the speedup values could not be obtained and hence plotted in Fig. 10.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Welling, W.S., Sheng, Y. & Zhu, M.M. CUDA-aware MPI implementation of Gibbs sampling for an IRT model. Cluster Comput 27, 1821–1830 (2024). https://doi.org/10.1007/s10586-023-04049-z
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10586-023-04049-z