Statistics and Computing

, Volume 27, Issue 5, pp 1365–1382 | Cite as

Fast parallel \(\alpha \)-stable distribution function evaluation and parameter estimation using OpenCL in GPGPUs

  • Guillermo Julián-Moreno
  • Jorge E. López de Vergara
  • Iván González
  • Luis de Pedro
  • Javier Royuela-del-Val
  • Federico Simmross-Wattenberg
Article
  • 1.6k Downloads

Abstract

\(\alpha \)-Stable distributions are a family of probability distributions found to be suitable to model many complex processes and phenomena in several research fields, such as medicine, physics, finance and networking, among others. However, the lack of closed expressions makes their evaluation analytically intractable, and alternative approaches are computationally expensive. Existing numerical programs are not fast enough for certain applications and do not make use of the parallel power of general purpose graphic processing units. In this paper, we develop novel parallel algorithms for the probability density function and cumulative distribution function—including a parallel Gauss–Kronrod quadrature—, quantile function, random number generator and maximum likelihood estimation of \(\alpha \)-stable distributions using OpenCL, achieving significant speedups and precision in all cases. Thanks to the use of OpenCL, we also evaluate the results of our library with different GPU architectures.

Keywords

Gaussian quadrature \(\alpha \)-Stable distribution Parallel algorithms Numerical algorithms OpenCL GPGPU 

References

  1. Achim, A., Bezerianos, A., Tsakalides, P.: Wavelet-based ultrasound image denoising using an alpha-stable prior probability model. In: Proceedings 2001 International Conference on Image Processing, vol. 2, pp. 221–224. IEEE (2001)Google Scholar
  2. Alpha Data. ADM-PCIE-7V3 datasheet, Nov 2013. URL http://www.alpha-data.com/pdfs/adm-pcie-7v3.pdf
  3. AMD. AMD graphics cores next (GCN) architecture. http://www.amd.com/Documents/GCN_Architecture_whitepaper.pdf (2012)
  4. AMD. AMD Radeon R9 series graphics cards. http://www.amd.com/en-us/products/graphics/desktop/r9#, 2013Google Scholar
  5. Arumugam, K., Godunov, A., Ranjan, D., Terzic, B., Zubair, M.: An efficient deterministic parallel algorithm for adaptive multidimensional numerical integration on GPUs. In: International Conference on Parallel Processing—The 42nd Annual Conference (ICPP 2013), pp. 486–491. IEEE, October 2013Google Scholar
  6. Bardou, F.: Lévy Statistics and Laser Cooling: how Rare Events Bring Atoms to Rest. Cambridge University Press, Cambridge (2002)MATHGoogle Scholar
  7. Belovas, I., et al.: Mixed-stable modeling of high-frequencyfinancial data: parallel computing approach. Technical report. Public center of BSU, Minsk (2013)Google Scholar
  8. DuMouchel, W.H.: On the asymptotic normality of the maximum-likelihood estimate when sampling from a stable distribution. Ann. Stat. 1(5), 948–957 (1973)MathSciNetCrossRefMATHGoogle Scholar
  9. Fang, J., Varbanescu, A.L., Sips, H.: A comprehensive performance comparison of CUDA and OpenCL. In: 2011 International Conference on Parallel Processing (ICPP), pp. 216–225. IEEE (2011)Google Scholar
  10. Giles, M.B.: Approximation of the inverse poisson cumulative distribution function. To appear in ACM Transactions of Mathematical Software (2015)Google Scholar
  11. Gnedenko, B.V., Kolmogorov, A.N.: Limit Distributions for Sums of Independent Random Variables. Series in Statistics, revised edn. Addison-Wesley, Cambridge, MA (1968)Google Scholar
  12. Gough, B.: GNU Scientific Library Reference Manual, 3rd edn. Network Theory Ltd., Bristol (2009)Google Scholar
  13. Hesterman, J.Y., Caucci, L., Kupinski, M.A., Barrett, H.H., Furenlid, L.R.: Maximum-likelihood estimation with a contracting-grid search algorithm. IEEE Trans. Nucl. Sci. 57(3), 1077–1084 (2010)CrossRefGoogle Scholar
  14. Koblents, E., Mguez, J., Rodrguez, M.A., Schmidt, A.M.: A nonlinear population monte carlo scheme for the bayesian estimation of parameters of \(\alpha \)-stable distributions. Comput. Stat. & Data Anal. 95, 57–74 (2016)MathSciNetCrossRefGoogle Scholar
  15. Koutrouvelis, I.A.: An iterative procedure for the estimation of the parameters of stable laws: an iterative procedure for the estimation. Commun. Stat.-Simul. Comput. 10(1), 17–28 (1981)MathSciNetCrossRefMATHGoogle Scholar
  16. Kronrod, A.S.: Nodes and weights of quadrature formulas: sixteen-place tables. Consultants Bureau (1965)Google Scholar
  17. Li, R., Zhao, Z., Qi, C., Zhou, X., Zhou, Y., Zhang, H.: Understanding the traffic nature of mobileinstantaneous messaging in cellular networks: a revisiting to \(\alpha \)-stable models. Access, IEEE 3, 1416–1422 (2015)Google Scholar
  18. Liang, Y., Chen, W.: A survey on computing lévy stable distributions and a new matlab toolbox. Signal Process. 93(1), 242–251 (2013)CrossRefGoogle Scholar
  19. Lombardi, M.J.: Bayesian inference for \(\alpha \)-stable distributions: a random walk MCMC approach. Comput. Stat. & Data Anal. 51(5), 2688–2700 (2007)MathSciNetCrossRefMATHGoogle Scholar
  20. McCulloch, J.H.: Simple consistent estimators of stable distribution parameters. Commun. Stat.-Simul. Comput. 15(4), 1109–1136 (1986)MathSciNetCrossRefMATHGoogle Scholar
  21. Menn, C., Rachev, S.T.: Calibrated fft-based density approximations for \(\alpha \)-stable distributions. Comput. Stat. & Data Anal. 50(8), 1891–1904 (2006)MathSciNetCrossRefMATHGoogle Scholar
  22. Mittnik, S., Rachev, S.T.: Modeling asset returns with alternative stable distributions. Econom. Rev. 12(3), 261–330 (1993)MathSciNetCrossRefMATHGoogle Scholar
  23. Nolan, J.P.: Numerical calculation of stable densities and distribution functions. Commun. Stat. Stoch. Models 13(4), 759–774 (1997)MathSciNetCrossRefMATHGoogle Scholar
  24. Nolan, J.P.: Stable Distributions—models for Heavy Tailed Data. Birkhauser, Boston, MA (2015). In Progress, Chap 1. http://academic2.american.edu/~jpnolan
  25. NVIDIA. NVIDIA OpenCL Best Practices Guide, August 2009aGoogle Scholar
  26. NVIDIA. Whitepaper: NVIDIA’s next generation CUDA compute architecture. Fermi (2009b). http://www.nvidia.com/content/PDF/fermi_white_papers/NVIDIA_Fermi_Compute_Architecture_Whitepaper.pdf
  27. NVIDIA. Tesla M2090 dual-slot computing processor module board spec, June 2012Google Scholar
  28. NVIDIA. Tesla K40 GPU active accelerator board spec, Nov 2013Google Scholar
  29. NVIDIA. Whitepaper: NVIDIA’s next generation CUDA compute architecture. Kepler GK110/210 (2014). http://international.download.nvidia.com/pdf/kepler/NVIDIA-Kepler-GK110-GK210-Architecture-Whitepaper.pdf
  30. Robinson, G.K.: Practical computing for finite moment log-stable distributions to model financial risk. Stat. Comput. 25(6), 1233–1246 (2014)MathSciNetCrossRefMATHGoogle Scholar
  31. Royuela-del-Val, J., Simmross-Wattenberg, F., Alberola-López, C.: Libstable: fast, parallel and high-precision computation of \(\alpha \)-stable distributions in C/C++ and MATLAB. J. Stat. Softw. .https://uvadoc.uva.es/bitstream/10324/15155/1/RoyuelaJSSoft.pdf (2016)
  32. Salas-González, D., Górriz, J.M., Ramírez, J., Schloegl, M., Lang, E.W., Ortiz, A.: Parameterization of the distribution of white and grey matter in MRI using the \(\alpha \)-stable distribution. Comput. Biol. Med. 43(5), 559–567 (2013)CrossRefGoogle Scholar
  33. Simmross-Wattenberg, F., Asensio-Pérez, J.I., Casaseca-de-la Higuera, P., Martín-Fernández, M., Dimitriadis, I., Alberola-López, C.: Anomaly detection in network traffic based on statistical inference and alpha-stable modeling. IEEE Trans. Dependable Secur. Comput. 8(4), 494–509 (2011)CrossRefGoogle Scholar
  34. Simmross-Wattenberg, F., Martín-Fernández, M., Higuera, P.C.D., Alberola-López, C.: Fast calculation of \(\alpha \)-stable density functions based on off-line precomputations. Application to ml parameter estimation. Digit. Signal Process. 38, 1–12 (2015)MathSciNetCrossRefGoogle Scholar
  35. Stone, J.E., Gohara, D., Shi, G.: OpenCL: a parallel programming standard for heterogeneous computing systems. Comput. Sci. & Eng. 12(1–3), 66–73 (2010)CrossRefGoogle Scholar
  36. Thuerck, D., Widmer, S., Kuijper, A., Goesele, M.: Efficient heuristic adaptive quadrature on GPUs: design and evaluation. In: Parallel Processing and Applied Mathematics, pp. 652–662. Springer, Berlin (2014)Google Scholar
  37. Weron, A., Weron, R.: Computer Simulation of Lévy \(\alpha \)-stable Variables and Processes. Springer, New York (1995)MATHGoogle Scholar
  38. Wuertz, D., Maechler, M.: CRAN Package ’stabledist’. https://cran.r-project.org/web/packages/stabledist/stabledist.pdf (2015)

Copyright information

© Springer Science+Business Media New York 2016

Authors and Affiliations

  • Guillermo Julián-Moreno
    • 1
  • Jorge E. López de Vergara
    • 1
  • Iván González
    • 1
  • Luis de Pedro
    • 1
  • Javier Royuela-del-Val
    • 2
  • Federico Simmross-Wattenberg
    • 2
  1. 1.Department of Electronics and Communication Technologies, Escuela Politécnica SuperiorUniversidad Autónoma de MadridMadridSpain
  2. 2.Image Processing Lab, E.T.S.I. TelecomunicaciónUniversidad de ValladolidValladolidSpain

Personalised recommendations