A Scalable Approach for LRT Computation in GPGPU Environments

  • Linsey Xiaolin Pang
  • Sanjay Chawla
  • Bernhard Scholz
  • Georgina Wilcox
Part of the Lecture Notes in Computer Science book series (LNCS, volume 7808)


In this paper we propose new algorithmic techniques for massively data parallel computation of the Likelihood Ratio Test (LRT) on a large spatial data grid. LRT is the state-of-the-art method for identifying hotspots or anomalous regions in spatially referenced data. LRT is highly adaptable permitting the use of a large class of statistical distributions to model the data. However, standard sequential implementations of LRT may take several days on modern machines to identify anomalous regions even for moderately sized spatial grids.

This work claims three novel contributions. First, we devise a dynamic program with a pre-processing step of \(\mathcal O(n^2)\) that allows us to compute the statistic for any given region in \(\mathcal O(1)\), where n is the length of the grid. Second, we propose a scheme to accelerate the likelihood computation of a complement region using a bounding technique. Third, we provide a parallelization strategy for the LRT computation on GPGPUs. In concert all three contributions result in a speed up of nearly four hundred times reducing the LRT computation time of large spatial grids from several days to minutes.


Spatial outlier GPGPUs LRT 1EXP upper-bounding 


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
  2. 2.
  3. 3.
    Agarwal, D., Phillips, J.M., Venkatasubramanian, S.: The hunting of the bump: On maximizing statistical discrepancy. In: SODA, pp. 1137–1146 (2006)Google Scholar
  4. 4.
    Beutel, A., Mølhave, T., Agarwal, P.K.: Natural neighbor interpolation based grid dem construction using a gpu. In: GIS 2010, pp. 172–181. ACM, New York (2010)Google Scholar
  5. 5.
    Gregerson, A.: Implementing fast mri gridding on gpus via. cuda. Nvidia Tech. Report on Medical Imaging using CUDA (2008)Google Scholar
  6. 6.
    Hong, S., Kim, S.K., Oguntebiv, T., Olukotun, K.: Efficient parallel graph exploration on multi-core cpu and gpu. In: Proceedings of the 16th ACM Symposium on Principles and Practice of Parallel Programming, PPoPP 2011 (2011)Google Scholar
  7. 7.
    Larew, S.G., Maciejewski, R., Woo, I., Ebert, D.S.: Spatial scan statistics on the gpgpu. In: Proceedings of the Visual Analytics in Healthcare Workshop at the IEEE Visualization Conference (2010)Google Scholar
  8. 8.
    Wu, M., Song, X., Jermaine, C., Ranka, S., Gums, J.: A LRT Framework for Fast Spatial Anomaly Detection. In: Proceedings of the 15th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD 2009), pp. 887–896 (2009)Google Scholar
  9. 9.
    Pang, L.X., Chawla, S., Liu, W., Zheng, Y.: On mining anomalous patterns in road traffic streams. In: Tang, J., King, I., Chen, L., Wang, J. (eds.) ADMA 2011, Part II. LNCS, vol. 7121, pp. 237–251. Springer, Heidelberg (2011)CrossRefGoogle Scholar
  10. 10.
    Wilks, S.S.: The large sample distribution of the likelihood ratio for testing composite hypotheses. Annals of Mathematical Statistics (9), 60–62 (1938)Google Scholar
  11. 11.
    Vuduc, R., Chandramowlishwaranv, A., Choi, J., Guney, M., Shringarpure, A.: On the limits of gpu acceleration. In: HotPar 2010 Proceedings of the 2nd USENIX Conference on Hot Topics in Parallelism, pp. 237–251 (2010)Google Scholar
  12. 12.
    Wu, R., Zhang, B., Hsu, M.: Gpu-accelerated large scale analytics. In: IACM UCHPC 2009: Second Workshop on UnConventional High Performance Computing (2009)Google Scholar
  13. 13.
    Wu, R., Zhang, B., Hsu, M.C.: Clustering billions of data points using gpus. In: IACM UCHPC 2009: Second Workshop on UnConventional High Performance Computing (2009)Google Scholar
  14. 14.
    Zhao, S.S., Zhou, C.: Accelerating spatial clustering detection of epidemic disease with graphics processing unit. In: Proceedings of Geoinformatics, pp. 1–6 (2010)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2013

Authors and Affiliations

  • Linsey Xiaolin Pang
    • 1
    • 2
  • Sanjay Chawla
    • 1
  • Bernhard Scholz
    • 1
  • Georgina Wilcox
    • 1
  1. 1.School of Information TechnologiesUniversity of SydneyAustralia
  2. 2.NICTASydneyAustralia

Personalised recommendations