Enhanced SMC$^2$: Leveraging Gradient Information from Differentiable Particle Filters Within Langevin Proposals
Abstract
Sequential Monte Carlo Squared (SMC) 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 SMC 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 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.
Keywords
Cite
@article{arxiv.2407.17296,
title = {Enhanced SMC$^2$: Leveraging Gradient Information from Differentiable Particle Filters Within Langevin Proposals},
author = {Conor Rosato and Joshua Murphy and Alessandro Varsi and Paul Horridge and Simon Maskell},
journal= {arXiv preprint arXiv:2407.17296},
year = {2024}
}
Comments
8 pages, 3 images. Accepted to 2024 IEEE International Conference on Multisensor Fusion and Integration (MFI 2024). https://mfi2024.org/. arXiv admin note: text overlap with arXiv:2311.12973