Abstract:Parallel Markov Chain Monte Carlo (pMCMC) algorithms generate clouds of proposals at each step to efficiently resolve a target probability distribution. We build a rigorous foundational framework for pMCMC algorithms that situates these methods within a unified 'extended phase space' measure-theoretic formalism. Drawing on our recent work that provides a comprehensive theory for reversible single proposal methods, we herein derive general criteria for multiproposal acceptance mechanisms which yield ergodic chains on general state spaces. Our formulation encompasses a variety of methodologies, including proposal cloud resampling and Hamiltonian methods, while providing a basis for the derivation of novel algorithms. In particular, we obtain a top-down picture for a class of methods arising from 'conditionally independent' proposal structures. As an immediate application, we identify several new algorithms including a multiproposal version of the popular preconditioned Crank-Nicolson (pCN) sampler suitable for high- and infinite-dimensional target measures which are absolutely continuous with respect to a Gaussian base measure. To supplement our theoretical results, we carry out a selection of numerical case studies that evaluate the efficacy of these novel algorithms. First, noting that the true potential of pMCMC algorithms arises from their natural parallelizability, we provide a limited parallelization study using TensorFlow and a graphics processing unit to scale pMCMC algorithms that leverage as many as 100k proposals at each step. Second, we use our multiproposal pCN algorithm (mpCN) to resolve a selection of problems in Bayesian statistical inversion for partial differential equations motivated by fluid measurement. These examples provide preliminary evidence of the efficacy of mpCN for high-dimensional target distributions featuring complex geometries and multimodal structures.

CosmoHammer: Cosmological parameter estimation with the MCMC Hammer

Hamiltonian Monte Carlo reconstruction from peculiar velocities

Bilby-MCMC: An MCMC sampler for gravitational-wave inference

A Single SMC Sampler on MPI that Outperforms a Single MCMC Sampler

The future of cosmological likelihood-based inference: accelerated high-dimensional parameter estimation and model comparison

A comparison of Bayesian sampling algorithms for high-dimensional particle physics and cosmology applications

A complete framework for cosmological emulation and inference with CosmoPower

CMBAnalysis: A Modern Framework for High-Precision Cosmic Microwave Background Analysis

CosmoPower-JAX: high-dimensional Bayesian inference with differentiable cosmological emulators

Cosmoglobe: Towards end-to-end CMB cosmological parameter estimation without likelihood approximations

Field-Level Inference with Microcanonical Langevin Monte Carlo

HACC: Simulating Sky Surveys on State-of-the-Art Supercomputing Architectures

A Serverless Architecture for Efficient and Scalable Monte Carlo Markov Chain Computation

Learn-As-You-Go Acceleration of Cosmological Parameter Estimates

Faster cosmological analysis with power spectrum without simulations

Markov Chain Monte Carlo methods applied to measuring the fine structure constant from quasar spectroscopy

Parallel MCMC Algorithms: Theoretical Foundations, Algorithm Design, Case Studies

21cmfish: Fisher-matrix framework for fast parameter forecasts from the cosmic 21-cm signal

Almanac: MCMC-based signal extraction of power spectra and maps on the sphere

Capse.jl: efficient and auto-differentiable CMB power spectra emulation

Particle MCMC algorithms and architectures for accelerating inference in state-space models