Gibbs sampler optimization for analysis of a granulated medium
A new variant of the method of probability density distribution recovery for solving topical modeling problems is described. Disadvantages of the Gibbs sampling algorithm are considered, and a modified variant, called the “granulated sampling method,” is proposed. Based on the results of statistical modeling, it is shown that the proposed algorithm is characterized by higher stability as compared to other variants of Gibbs sampling.
Unable to display preview. Download preview PDF.
- 2.I. Chernyavsky, T. Alexandrov, P. Maass, and S. Nikolenko, Proceedings of the German Conference on Bioinformatics (September, 2012), pp. 39–48.Google Scholar
- 3.Handbook of Markov Chain Monte Carlo, Ed. by S. Brooks, A. Gelman, G. L. Jones, and X.-L. Meng (Chapman & Hall/CRC Press, 2011), pp. 383–399.Google Scholar
- 4.B. A. Berg and A. Billoire, Markov Chain Monte Carlo Simulations (John Wiley & Sons, 2008).Google Scholar
- 6.S. Bodrunova, S. Koltsov, O. Koltsova, S. Nikolenko, and A. Shimorina, Proceedings of the 12th Mexican Int. Conf. on Artificial Intelligence (MICAI 2013) (Springer Verlag, Berlin, 2013), Part I, pp. 265–274.Google Scholar
- 7.D. Blei, A. Ng, and M. Jordan (Ed. by J. Lafferty), J. Machine Learn. Res. 3, 993 (2003).Google Scholar
- 9.C. Nelson et al., Proceedings of the IEEE Conference on Frequency Control (2012).Google Scholar
- 12.S. Koltsov, O. Koltsova, and S. Nikolenko, Proceedings of the ACM Web Science Conference (WebSci’14, June 23–26, 2014, Bloomington, IN, USA), pp. 161–165.Google Scholar