A Comparison of Kernels for ABC-SMC

2025/11/09 by Dennis Prangle, Cecilia Viscardi, Prangle, Dennis +3 · 2 voices
Computer Science · Mathematics · #Computation (stat.CO) #FOS: Computer and information sciences #Gaussian Processes and Bayesian Inference #Markov Chains and Monte Carlo Methods #Target Tracking and Data Fusion in Sensor Networks #stat.CO

paper · pdf · doi:10.48550/arxiv.2511.06351

openalex publication_date 2025/11/09 · arxiv published 2025/11/09 · arxiv updated 2025/11/09 · openalex created_date 2025/11/12 · openalex updated_date 2026/07/28

Abstract

A popular method for likelihood-free inference is approximate Bayesian computation sequential Monte Carlo (ABC-SMC) algorithms. These approximate the posterior using a population of particles, which are updated using Markov kernels. Several such kernels have been proposed. In this paper we review these, highlighting some less well known choices, and proposing some novel options. Further, we conduct an extensive empirical comparison of kernel choices. Our results suggest using a one-hit kernel with a mixture proposal as a default choice.

Citations

• The Statistical Accuracy of Neural Posterior and Likelihood Estimation
• Continual Repeated Annealed Flow Transport Monte Carlo
• A Comparison of Likelihood-Free Methods With and Without Summary Statistics
• Annealed Flow Transport Monte Carlo
• Benchmarking Simulation-Based Inference
• Waste-free Sequential Monte Carlo
• A Review of Multi‐Compartment Infectious Disease Models
• Distilling Importance Sampling for Likelihood Free Inference
• Neural Spline Flows
• Sequential Neural Likelihood: Fast Likelihood-free Inference with Autoregressive Flows
• ABC Samplers
• Delayed acceptance ABC-SMC
• Negative association, ordering and convergence of resampling methods
• Adapting the ABC distance function
• Robust Bayesian inference via coarsening
• Accelerating Metropolis-Hastings algorithms by Delayed Acceptance
• On Bayesian inference for the M/G/1 queue with efficient MCMC sampling
• On the Convergence of Adaptive Sequential Monte Carlo Methods
• On optimality of kernels for approximate Bayesian computation using\n sequential Monte Carlo
• On the Stability of Sequential Monte Carlo Methods in High Dimensions
• Sequential Monte Carlo samplers: error bounds and insensitivity to initial conditions
• Sequential monte carlo samplers
• Approximate Bayesian computational methods

Discussions

• New paper: arxiv.org/abs/2511.06351. We compare different MCMC kernels for ABC-SMC, and recommend a one- hit kernel with mixture proposal. (Repost with corrected link!) [bsky, 3 points, 0 comments]
• Dennis Prangle, Cecilia Viscardi, Sammy Ragy: A Comparison of Kernels for ABC-SMC https://arxiv.org/abs/2511.06351 https://arxiv.org/pdf/2511.06351 https://arxiv.org/html/2511.06351 [bsky, 0 points, 0 comments]

Related

• On optimality of kernels for approximate Bayesian computation using\n sequential Monte Carlo
• Properties of Marginal Sequential Monte Carlo Methods
• Delayed acceptance ABC-SMC
• An Adaptive Sequential Monte Carlo Sampler
• Approximate Bayesian Computation via Population Monte Carlo and Classification