Discontinuous Hamiltonian Monte Carlo for discrete parameters and discontinuous likelihoods

  title={Discontinuous Hamiltonian Monte Carlo for discrete parameters and discontinuous likelihoods},
  author={A. Nishimura and D. Dunson and J. Lu},
  journal={arXiv: Computation},
  • A. Nishimura, D. Dunson, J. Lu
  • Published 2017
  • Mathematics
  • arXiv: Computation
  • Hamiltonian Monte Carlo has emerged as a standard tool for posterior computation. In this article, we present an extension that can efficiently explore target distributions with discontinuous densities. Our extension in particular enables efficient sampling from ordinal parameters though embedding of probability mass functions into continuous spaces. We motivate our approach through a theory of discontinuous Hamiltonian dynamics and develop a corresponding numerical solver. The proposed solver… CONTINUE READING

    Figures and Tables from this paper.

    Paper Mentions

    Quantum-Inspired Hamiltonian Monte Carlo for Bayesian Sampling
    • 1
    • PDF
    Discrete Sampling using Semigradient-based Product Mixtures
    • 1
    • PDF
    Improving Predictive Uncertainty Estimation using Dropout-Hamiltonian


    Publications referenced by this paper.
    MCMC Using Hamiltonian Dynamics
    • 1,623
    • PDF
    Riemann manifold Langevin and Hamiltonian Monte Carlo methods
    • 1,013
    • PDF
    Practical Markov Chain Monte Carlo
    • 1,575
    Geometric Numerical Integration: Structure Preserving Algorithms for Ordinary Differential Equations
    • 2,752
    • Highly Influential
    • PDF
    An adaptive Metropolis algorithm
    • 1,752
    • Highly Influential
    • PDF
    The No-U-turn sampler: adaptively setting path lengths in Hamiltonian Monte Carlo
    • 1,645
    • Highly Influential
    • PDF
    Weak convergence and optimal scaling of random walk Metropolis algorithms
    • 1,398
    • Highly Influential
    • PDF
    CODA: convergence diagnosis and output analysis for MCMC
    • 2,231
    Hybrid Monte Carlo
    • 1,893