Advertisement

A General Metric for Riemannian Manifold Hamiltonian Monte Carlo

  • Michael Betancourt
Part of the Lecture Notes in Computer Science book series (LNCS, volume 8085)

Abstract

Markov Chain Monte Carlo (MCMC) is an invaluable means of inference with complicated models, and Hamiltonian Monte Carlo, in particular Riemannian Manifold Hamiltonian Monte Carlo (RMHMC), has demonstrated success in many challenging problems. Current RMHMC implementations, however, rely on a Riemannian metric that limits their application. In this paper I propose a new metric for RMHMC without these limitations and verify its success on a distribution that emulates many hierarchical and latent models.

Keywords

Riemannian Manifold Markov Chain Monte Carlo Target Distribution Hamiltonian Evolution Euclidean Manifold 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Aizu, K.: Parameter differentiation of quantum-mechanical linear operators. Journal of Mathematical Physics 4(6), 762–775 (1963)MathSciNetzbMATHCrossRefGoogle Scholar
  2. 2.
    Amari, S., Nagaoka, H.: Methods of information geometry, vol. 191. Amer. Mathematical Society (2007)Google Scholar
  3. 3.
    Betancourt, M., Stein, L.C.: The Geometry of Hamiltonian Monte Carlo (2011)Google Scholar
  4. 4.
    Celis, M., Dennis, J.E.: A., T.R.: A trust region strategy for nonlinear equality constrained optimization. In: Boggs, P., Byrd, R., Schnabel, R. (eds.) Numerical Optimization 1984, SIAM, Philadelphia (1985)Google Scholar
  5. 5.
    Duane, S., Kennedy, A., Pendleton, B.J., Roweth, D.: Hybrid monte carlo. Physics Letters B 195(2), 216–222 (1987)CrossRefGoogle Scholar
  6. 6.
    Gelman, A., Carlin, J., Stern, H., Rubin, D.: Bayesian Data Analysis. Chapman & Hall/CRC Press, Boca Raton (2004)zbMATHGoogle Scholar
  7. 7.
    Geyer, C.: Introduction to mcmc. In: Brooks, S., Gelman, A., Jones, G.L., Meng, X.L. (eds.) Handbook of MCMC. CRC Press, New York (2011)Google Scholar
  8. 8.
    Girolami, M., Calderhead, B.: Riemann manifold langevin and hamiltonian monte carlo methods. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 73(2), 123–214 (2011)MathSciNetCrossRefGoogle Scholar
  9. 9.
    Hoffman, M.D., Gelman, A.: The No-U-Turn Sampler: Adaptively Setting Path Lengths in Hamiltonian Monte Carlo. ArXiv e-prints (November 2011)Google Scholar
  10. 10.
    Magnus, J., Neudecker, H.: Matrix Differential Calculus with Applications in Statistics and Econometrics. Wiley, New York (2007)Google Scholar
  11. 11.
    Moler, C., Van Loan, C.: Nineteen dubious ways to compute the exponential of a matrix, twenty-five years later. SIAM Review 45(1), 3–49 (2003)MathSciNetzbMATHCrossRefGoogle Scholar
  12. 12.
    Murray, I., Prescott Adams, R.: Slice Sampling Covariance Hyperparameters of Latent Gaussian Models. ArXiv e-prints (June 2010)Google Scholar
  13. 13.
    Neal, R.: Slice sampling. Annals of Statistics, 705–741 (2003)Google Scholar
  14. 14.
    Neal, R.: Mcmc using hamiltonian dynamics. In: Brooks, S., Gelman, A., Jones, G.L., Meng, X.L. (eds.) Handbook of MCMC. CRC Press, New York (2011)Google Scholar
  15. 15.
    Spivak, M.: A Comprehensive Introduction to Differential Geometry, vol. 1. Publish or Perish, Inc., Houston (2005)Google Scholar
  16. 16.
    Wilcox, R.M.: Exponential operators and parameter differentiation in quantum physics. Journal of Mathematical Physics 8(4), 962–982 (1967)MathSciNetzbMATHCrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2013

Authors and Affiliations

  • Michael Betancourt
    • 1
  1. 1.Department of StatisticsColumbia UniversityNew YorkUSA

Personalised recommendations