Diffusion differentiable resampling
This addresses the need for differentiable resampling in particle filtering, which is incremental as it improves upon existing methods for specific applications like parameter estimation and high-dimensional image processing.
The paper tackled the problem of differentiable resampling in sequential Monte Carlo methods by proposing a new informative resampling method based on an ensemble score diffusion model, which outperformed state-of-the-art differentiable resampling methods on multiple benchmarks and achieved competitive performance in learning a complex dynamics-decoder model with high-dimensional image observations.
This paper is concerned with differentiable resampling in the context of sequential Monte Carlo (e.g., particle filtering). We propose a new informative resampling method that is instantly differentiable, based on an ensemble score diffusion model. We theoretically prove that our diffusion resampling method provides a consistent resampling distribution, and we show empirically that it outperforms the state-of-the-art differentiable resampling methods on multiple filtering and parameter estimation benchmarks. Finally, we show that it achieves competitive end-to-end performance when used in learning a complex dynamics-decoder model with high-dimensional image observations.