Abstract: “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 through …

This was a particularly bad paper with fatal methodological flaws that had been shopped around journals for a long time. I’m not sure the current state of the paper but I also can’t review it behind a paywall. Discontinuities in the target density function cause problems for the numerical integrat…

Maybe also @betanalpha could be asked to provide an up-to-date account on the state of the art when it comes to HMC for discrete spaces (or point to something he’s written).

I realize this is a somewhat old thread. Since I happened to come across this thread for the first time, though, I wanted to respond to @betanalpha ’s unfair critique of our work. with fatal methodological flaws Besides the theory developed in our paper, the correctness of our method is empirical…

Discontinuous Hamiltonian Monte Carlo for discrete parameters and discontinuous likelihoods

Developers

maxbiostat March 30, 2020, 7:37pm 2

Maybe this discussion would be helpful.

Topic		Replies	Views
What (exactly) happens for only piecewise continuous posterior densities, gradients or wrong gradients? Algorithms	4	1384	April 26, 2022
How Hamiltonian Monte Carlo handle the discontinuity at zero in hurdle models? Modeling techniques	6	769	June 14, 2021
A possible way to sample from discrete variables? Modeling	3	844	July 6, 2017
Can I do HMC with the wrong Hamiltonian? Algorithms	4	681	May 1, 2020
How to write a logical Matrix in the parameter chunk? Modeling	3	481	January 29, 2019

Discontinuous Hamiltonian Monte Carlo for discrete parameters and discontinuous likelihoods

Related topics