On this website I aim to collect references to papers on the use and analysis of Piecewise Deterministic Markov Processes (PDMPs) for use in MCMC. I will try to list all papers that I find relevant, will put years of first appearance on arXiv or elsewhere, and list in reverse chronological order. Inevitably this page will reflect some of my own judgements. If you want to bring some paper to my attention, or feel I should correct something, please write me at `joris.bierkens AT tudelft.nl`

.

Markov Chain Monte Carlo (MCMC) is a probabilistic computational method which is of vital importance in data science (statistics, machine learning) as well as in the physical sciences (physics, chemistry, biology). For example, MCMC is applied in Bayesian statistics, for the training of (deep) neural networks, as well as for the simulation of many-particle systems.

In recent years it has been established that PDMPs may play a very useful role in designing new efficient MCMC methods which have good convergence properties and allow various clever tricks in dealing with large data sets (subsampling of the data, for example).

The following images are generated using the R package `RZigZag`

, which can be installed via the command `install.packages("RZigZag")`

. The plots can then be produced using this R script.

What we see are two-dimensional projections of higher dimensional trajectories of two PDMP samplers: the Zig-Zag Sampler and the Bouncy Particle Sampler (BPS). The trajectories can be seen to be continuous (in contrast to most MCMC algorithms). Also shown are discrete samples drawn along the trajectories, which can be interpreted in the classical MCMC sense. Both plots in this example correspond to a 5-dimensional standard normal distribution. The BPS trajectory is generated using a refreshment rate of `0.01`

.

Here papers are listed which concern the mathematical analysis of the stochastic processes underlying PDMP based algorithms.

- C. Andrieu, A. Durmus, N. Nüsken, J. Roussel, Hypercoercivity of Piecewise Deterministic Markov Process-Monte Carlo (2018).
*This important paper gives results in a very general context for scaling with dimension of various PDMP samplers. The obtained results seem to be sharp, for example in the sense that they agree with the results obtained for the specialized case of a factorized distribution in Bierkens, Kamatani and Roberts (2018) discussed below. No doubt the title will soon to change to "Hypocoercivity ... "* - G. Deligiannidis, D. Paulin, A. Doucet, Randomized Hamiltonian Monte Carlo as Scaling Limit of the Bouncy Particle Sampler and Dimension-Free Convergence Rates (2018).
*For a constant refreshment rate (relative to standard normal velocities in stationarity), the first coordinate of the BPS has a scaling limit as the dimension d goes to infinity, implying that the mixing time of the first coordinate process is dimension-independent. Note however that in Bierkens, Kamatani and Roberts (2018) we show that in order for the radial component to mix we require the refreshment rate to grow at the same speed as (the norm of) the velocity.* - J. Bierkens, K. Kamatani, G. Roberts, High-dimensional scaling limits of piecewise deterministic sampling algorithms (2018).
*In the case of a standard normal distribution on R^d, the computational efficiency of Zig-Zag and BPS (with constant positive refreshment) are compared by means of scaling limits for certain summary statistics.* - A. Durmus, A. Guillin, P. Monmarché, Geometric ergodicity of the bouncy particle sampler (2018).
*Compared to Deligiannidis et al. (2017) the conditions for exponential ergodicity of BPS are weakened, based on a coupling argument.* - J. Bierkens, G. Roberts, P.-A. Zitt, Ergodicity of the zigzag process (2017).
*In this paper it is shown that under mild conditions the zigzag process is (exponentially) ergodic, even without refreshments of the velocity.* - G. Deligiannidis, A. Bouchard-Côté, A. Doucet, Exponential Ergodicity of the Bouncy Particle Sampler (2017)
- J. Bierkens, Andrew Duncan, Limit theorems for the Zig-Zag process (2016)
- J. Bierkens, Gareth Roberts, A piecewise deterministic scaling limit of Lifted Metropolis-Hastings in the Curie-Weiss model (2015).
*In this work we obtained the zigzag process as a scaling limit of Lifted Metropolis-Hastings, a discrete non-reversible MCMC algorithm (see the paper by Turitsyn et al. (2008) listed below). We obtain exponential ergodicity of the Zig-Zag process in the one-dimensional case.* - J. Fontbona, H. Guérin, F. Malrieu, Long time behavior of telegraph processes under convex potentials (2015).
*Analysis of a one-dimensional process with piecewise linear dynamics and space dependent switching rate for a chemotaxis model, essentially what I call the one-dimensional zigzag process.* - P. Monmarché, Piecewise deterministic simulated annealing (2014).
*Most of the building blocks for sampling in the one-dimensional case are here (in a later version, this is extended to multiple dimensions). The focus is on using PDMPs for optimization by means of simulated annealing.* - P. Monmarché, Étude spectrale minutieuse de processus moins indécis que les autres (2012).
*Spectral analysis of a one-dimensional piecewise linear Markov process with constant switching intensity on the torus; English summary here.* - J. Fontbona, H. Guérin, F. Malrieu, Quantitative estimates for the long time behavior of an ergodic variant of the telegraph process
(2010).
*Analysis of a one-dimensional process with piecewise linear dynamics and constant switching rate for a chemotaxis model.* - M. Kac, A stochastic model related to the telegrapher's equation (1974).
*Further analysis of Goldstein's telegraph process.* - S. Goldstein, On diffusion by discontinuous movements, and on the telegraph equation (1951).
*Analysis of a piecewise deterministic linear process with constant switching intensity.*

Here papers are listed which describe sampling methods based on PDMPs.

- C. Wu, C. Robert, The Coordinate Sampler: A Non-Reversible Gibbs-like MCMC Sampler (2018)
- P. Vanetti, A. Bouchard-Côté, George Deligiannidis, Arnaud Doucet, Piecewise Deterministic Markov Chain Monte Carlo (2017)
- C. Wu, C. Robert, Generalized Bouncy Particle Sampler (2017)
- J. Bierkens, A. Bouchard-Côté, A. Doucet, A. Duncan, P. Fearnhead, T. Lienart, G. Roberts, S. Vollmer, Piecewise Deterministic Markov Processes for Scalable Monte Carlo on Restricted Domains (2017)
- P. Fearnhead, J. Bierkens, M. Pollock, G. Roberts, Piecewise Deterministic Markov Processes for Continuous-Time Monte Carlo (2016)
- J. Bierkens, P. Fearnhead, G. Roberts, The Zig-Zag Process and Super-Efficient Sampling for Bayesian Analysis of Big Data (2016)
- A. Bouchard-Côté, S. Vollmer, A. Doucet, The Bouncy Particle Sampler: A Non-Reversible Rejection-Free Markov Chain Monte Carlo Method (2015)

- M. Michel, S. Kapfer, W. Krauth, Generalized event-chain Monte Carlo: Constructing rejection-free global-balance algorithms from infinitesimal steps (2013)
- E. Peters, G. de With, Rejection-free Monte Carlo sampling for general potentials (2011)
- K. Turitsyn, M. Chertkov, M. Vucelja, Irreversible Monte Carlo Algorithms for Efficient Sampling (2008)

- A. Pakman, Binary Bouncy Particle Sampler (2017)
- A. Pakman, D. Gilboa, D. Carlson, L. Paninski, Stochastic Bouncy Particle Sampler (2016)

*Last update: February 2019*