Knots and variance ordering of sequential Monte Carlo algorithms

2025-10-02

AI Summary

Particle filters suffer from high variance and low efficiency when approximating intractable integrals in Bayesian inference. Method: This paper introduces the “node operator”—a novel variance-reduction technique that explicitly incorporates potential function information into the transition kernel to construct an improved sequential importance sampling mechanism. Grounded in the Feynman–Kac framework, it establishes a partial order on asymptotic variances across particle filters. Contribution/Results: For the first time, a minimal modification to the fully adapted particle filter guarantees strict variance ordering for all test functions. The theory unifies and generalizes existing approaches—including marginalization and auxiliary particle filtering—providing a provably sound, comparatively tractable paradigm for designing variance-controlled sequential Monte Carlo (SMC) algorithms. This yields significant improvements in both statistical efficiency and numerical stability.

Technology Category

• Reasoning under Uncertainty: Stochastic Optimization
• Machine Learning: Probabilistic Circuits and Graphical Models
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Security and Privacy: Privacy-enhancing technologies
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Sequential Monte Carlo algorithms, or particle filters, are widely used for approximating intractable integrals, particularly those arising in Bayesian inference and state-space models. We introduce a new variance reduction technique, the knot operator, which improves the efficiency of particle filters by incorporating potential function information into part, or all, of a transition kernel. The knot operator induces a partial ordering of Feynman-Kac models that implies an order on the asymptotic variance of particle filters, offering a new approach to algorithm design. We discuss connections to existing strategies for designing efficient particle filters, including model marginalisation. Our theory generalises such techniques and provides quantitative asymptotic variance ordering results. We revisit the fully-adapted (auxiliary) particle filter using our theory of knots to show how a small modification guarantees an asymptotic variance ordering for all relevant test functions.

Problem

• Improving particle filter efficiency with knot operators
• Ordering Feynman-Kac models by asymptotic variance
• Guaranteeing variance reduction in sequential Monte Carlo

Innovation

• Introduces knot operator for variance reduction
• Orders Feynman-Kac models by asymptotic variance
• Generalizes existing particle filter design techniques