Firefly Monte Carlo: Exact MCMC with Subsets of Data

Abstract

Markov chain Monte Carlo (MCMC) is a popular and successful general-purpose tool for Bayesian inference. However, MCMC cannot be practically applied to large data sets because of the prohibitive cost of evaluating every likelihood term at every iteration. Here we present Firefly Monte Carlo (FlyMC) an auxiliary variable MCMC algorithm that only queries the likelihoods of a potentially small subset of the data at each iteration yet simulates from the exact posterior distribution, in contrast to recent proposals that are approximate even in the asymptotic limit. FlyMC is compatible with a wide variety of modern MCMC algorithms, and only requires a lower bound on the per-datum likelihood factors. In experiments, we find that FlyMC generates samples from the posterior more than an order of magnitude faster than regular MCMC, opening up MCMC methods to larger datasets than were previously considered feasible.

References in corpus (3)

• The pseudo-marginal approach for efficient Monte Carlo computations
• Austerity in MCMC Land: Cutting the Metropolis-Hastings Budget
• Accelerating MCMC via Parallel Predictive Prefetching

Cited by in corpus (18)

• Stein Variational Gradient Descent: A General Purpose Bayesian Inference Algorithm
• Speeding Up MCMC by Efficient Data Subsampling
• Distributed ARIMA Models for Ultra-long Time Series
• Speeding Up MCMC by Delayed Acceptance and Data Subsampling
• A Survey of Bayesian Statistical Approaches for Big Data
• Hamiltonian Monte Carlo with Energy Conserving Subsampling
• AMAGOLD: Amortized Metropolis Adjustment for Efficient Stochastic Gradient MCMC
• No Free Lunch for Approximate MCMC
• The block-Poisson estimator for optimally tuned exact subsampling MCMC
• Asymptotically Optimal Exact Minibatch Metropolis-Hastings
• Sparse Variational Inference: Bayesian Coresets from Scratch
• A deterministic and computable Bernstein-von Mises theorem
• Mini-batch Metropolis-Hastings MCMC with Reversible SGLD Proposal
• Bayesian Coresets: Revisiting the Nonconvex Optimization Perspective
• Probabilistic Programs with Stochastic Conditioning
• An Algorithm for Distributed Bayesian Inference in Generalized Linear Models
• A Bayesian Hidden Semi-Markov Model with Covariate-Dependent State Duration Parameters for High-Frequency Data from Wearable Devices
• On MCMC for variationally sparse Gaussian processes: A pseudo-marginal approach