PyDPF: A Python Package for Differentiable Particle Filtering

2025-10-29

AI Summary

Traditional particle filters for state space models (SSMs) are non-differentiable, hindering end-to-end joint optimization of latent states and unknown parameters. To address this, we propose and implement a unified, differentiable particle filtering (DPF) framework in PyTorch. Our method ensures full gradient flow through the filtering pipeline by reformulating resampling operations—including SoftResample and differentiable multinomial resampling—while providing a standardized API that integrates multiple state-of-the-art DPF algorithms for flexible configuration and fair benchmarking. Experimental reproduction across standard SSM benchmarks demonstrates significant improvements in parameter estimation accuracy and training stability. The framework serves as an efficient, open-source, plug-and-play differentiable inference tool for modeling complex dynamical systems.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Intelligent Robots: State Estimation
• Reasoning under Uncertainty: Probabilistic Programming

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

State-space models (SSMs) are a widely used tool in time series analysis. In the complex systems that arise from real-world data, it is common to employ particle filtering (PF), an efficient Monte Carlo method for estimating the hidden state corresponding to a sequence of observations. Applying particle filtering requires specifying both the parametric form and the parameters of the system, which are often unknown and must be estimated. Gradient-based optimisation techniques cannot be applied directly to standard particle filters, as the filters themselves are not differentiable. However, several recently proposed methods modify the resampling step to make particle filtering differentiable. In this paper, we present an implementation of several such differentiable particle filters (DPFs) with a unified API built on the popular PyTorch framework. Our implementation makes these algorithms easily accessible to a broader research community and facilitates straightforward comparison between them. We validate our framework by reproducing experiments from several existing studies and demonstrate how DPFs can be applied to address several common challenges with state space modelling.

Problem

• Differentiable particle filters enable gradient-based parameter estimation
• Implementing unified API for multiple differentiable particle filtering algorithms
• Facilitating accessibility and comparison of state-space model optimization methods

Innovation

• Differentiable particle filters enable gradient-based optimization
• Unified PyTorch API for multiple differentiable particle filters
• Implementation facilitates algorithm comparison and broader accessibility