[
    {
        "id": "osp-16367",
        "type": "article-journal",
        "title": "Differentiable Systematic Resampling for Variational Sequential Monte Carlo",
        "author": [
            {
                "family": "Cumlin",
                "given": "Fredrik"
            },
            {
                "family": "Chatterjee",
                "given": "Saikat"
            }
        ],
        "URL": "https://omanscience.com/ar/articles/differentiable-systematic-resampling-for-variational-sequential-monte-carlo",
        "language": "en",
        "issued": {
            "date-parts": [
                [
                    2026
                ]
            ]
        },
        "abstract": "Particle filters are a standard tool for nonlinear state estimation, but their resampling step is discrete, preventing gradient-based learning in variational sequential Monte Carlo. We introduce Differentiable Systematic Resampling (DSR), a temperature-controlled relaxation of systematic resampling, that preserves the CDF-ordered, banded structure of systematic resampling while enabling full gradient flow. DSR converges to exact systematic resampling as the temperature vanishes, and we prove a pointwise exponential convergence rate for the induced bias. Compared to optimal-transport-based differentiable resampling, DSR avoids iterative solvers and has substantially lower computational overhead. Experiments on stochastic dynamical systems and real-world handwriting data show that DSR achieves comparable or superior filtering and dynamics learning performance."
    }
]