mlsanityData to 7 Oct 2026

Enhanced SMC2: Leveraging Gradient Information from Differentiable Particle Filters Within Langevin Proposals

arXiv 2407.17296 ↗First posted 24 Jul 2024stat.ML cs.LG stat.APcited by 4

Not on the map. Why?

Abstract (arXiv)

Sequential Monte Carlo Squared (SMC2) is a Bayesian method which can infer the states and parameters of non-linear, non-Gaussian state-space models. The standard random-walk proposal in SMC2 faces challenges, particularly with high-dimensional parameter spaces. This study outlines a novel approach by harnessing first-order gradients derived from a Common Random Numbers - Particle Filter (CRN-PF) using PyTorch. The resulting gradients can be leveraged within a Langevin proposal without accept/reject. Including Langevin dynamics within the proposal can result in a higher effective sample size and more accurate parameter estimates when compared with the random-walk. The resulting algorithm is parallelized on distributed memory using Message Passing Interface (MPI) and runs in O(log2N) time complexity. Utilizing 64 computational cores we obtain a 51x speed-up when compared to a single core. A GitHub link is given which provides access to the code.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Builds on 12 of its 40 references that are in the map. First posted on arXiv 24 Jul 2024.

Builds on

12 of 40 references.

References · 12

  1. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  2. 2018Particle Filter Networks with Application to Visual Localizationcited by 64
  3. 2021Differentiable Particle Filtering via Entropy-Regularized Optimal Transportcited by 59
  4. 2022Automatic Differentiation of Programs with Discrete Randomnesscited by 28
  5. 2023An overview of differentiable particle filters for data-adaptive sequential Bayesian inferencecited by 23
  6. 2021Differentiable Particle Filtering without Modifying the Forward Passcited by 18
  7. 2021Efficient Learning of the Parameters of Non-Linear Models using Differentiable Resampling in Particle Filterscited by 9
  8. 2015Getting Started with Particle Metropolis-Hastings for Inference in Nonlinear Dynamical Modelscited by 6
  9. 2013Particle Metropolis-Hastings using gradient and Hessian informationcited by 4
  10. 2022Particle-Based Score Estimation for State Space Model Learning in Autonomous Drivingcited by 3
  11. 2014Particle Metropolis adjusted Langevin algorithms for state space modelscited by 2
  12. 2024Revisiting semi-supervised training objectives for differentiable particle filterscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 3; 2026: 1.

4 papers.

  1. 2025Incorporating the ChEES Criterion into Sequential Monte Carlo Samplerscited by 3
  2. 2025Hess-MC2: Sequential Monte Carlo Squared using Hessian Information and Second Order Proposalscited by 1
  3. 2025Utilising Gradient-Based Proposals Within Sequential Monte Carlo Samplers for Training of Partial Bayesian Neural Networkscited by 0
  4. 2026Investigating Batch Inference in a Sequential Monte Carlo Framework for Neural Networkscited by 0