On Markov chain Monte Carlo methods for tall data

Abstract

Markov chain Monte Carlo methods are often deemed too computationally intensive to be of any practical use for big data applications, and in particular for inference on datasets containing a large number $n$ of individual data points, also known as tall datasets. In scenarios where data are assumed independent, various approaches to scale up the Metropolis-Hastings algorithm in a Bayesian inference context have been recently proposed in machine learning and computational statistics. These approaches can be grouped into two categories: divide-and-conquer approaches and, subsampling-based algorithms. The aims of this article are as follows. First, we present a comprehensive review of the existing literature, commenting on the underlying assumptions and theoretical guarantees of each method. Second, by leveraging our understanding of these limitations, we propose an original subsampling-based approach which samples from a distribution provably close to the posterior distribution of interest, yet can require less than $O(n)$ data point likelihood evaluations at each iteration for certain statistical models in favourable scenarios. Finally, we have only been able so far to propose subsampling-based methods which display good performance in scenarios where the Bernstein-von Mises approximation of the target posterior distribution is excellent. It remains an open challenge to develop such methods in scenarios where the Bernstein-von Mises approximation is poor.

References in corpus (12)

• An MCMC Approach to Classical Estimation
• The pseudo-marginal approach for efficient Monte Carlo computations
• Stochastic Gradient Hamiltonian Monte Carlo
• Fast Computation of Wasserstein Barycenters
• Austerity in MCMC Land: Cutting the Metropolis-Hastings Budget
• Speeding Up MCMC by Efficient Data Subsampling
• The Penalty Method for Random Walks with Uncertain Energies
• A Noisy Monte Carlo Algorithm
• Ergodicity of Approximate MCMC Chains with Applications to Large Data Sets
• The Fundamental Incompatibility of Hamiltonian Monte Carlo and Data Subsampling
• Consistency and fluctuations for stochastic gradient Langevin dynamics
• Unbiased Bayes for Big Data: Paths of Partial Posteriors

Cited by in corpus (28)

• Hands-on Bayesian Neural Networks -- a Tutorial for Deep Learning Users
• Speeding Up MCMC by Efficient Data Subsampling
• Fusion of Probability Density Functions
• Orthogonal parallel MCMC methods for sampling and optimization
• Expectation propagation as a way of life: A framework for Bayesian inference on partitioned data
• Piecewise Deterministic Markov Processes for Scalable Monte Carlo on Restricted Domains
• Leave Pima Indians alone: binary regression as a benchmark for Bayesian computation
• Speeding Up MCMC by Delayed Acceptance and Data Subsampling
• Ergodicity of Approximate MCMC Chains with Applications to Large Data Sets
• A Survey of Bayesian Statistical Approaches for Big Data
• Clock Monte Carlo methods
• Approximating Bayes in the 21st Century
• Propensity-to-Pay: Machine Learning for Estimating Prediction Uncertainty
• Theoretical properties of quasi-stationary Monte Carlo methods
• Computing Bayes: From Then 'Til Now'
• Fast Inference for Intractable Likelihood Problems using Variational Bayes
• A survey of Monte Carlo methods for noisy and costly densities with application to reinforcement learning and ABC
• Distributed Computation for Marginal Likelihood based Model Choice
• No Free Lunch for Approximate MCMC
• Average of Recentered Parallel MCMC for Big Data
• PAC-Bayes Bounds for Bandit Problems: A Survey and Experimental Comparison
• Mini-batch Metropolis-Hastings MCMC with Reversible SGLD Proposal
• Scalable Bayesian Nonparametric Clustering and Classification
• Stochastic Gradient MCMC with Stale Gradients
• Perturbation Analysis of Markov Chain Monte Carlo for Graphical Models
• Covariance Matrices for Mean Field Variational Bayes
• An Algorithm for Distributed Bayesian Inference in Generalized Linear Models
• Data Subsampling for Bayesian Neural Networks