Overview

Bayesian filtering studies how to update beliefs about hidden states as new observations arrive, which supports reliable prediction and decision-making when measurements are noisy or incomplete. Kalman and Gaussian filters offer efficient updates for roughly unimodal beliefs, while particle, ensemble, and variational methods handle nonlinear or non-Gaussian systems; newer approaches also use learned proposals, score models, and generative flows. These methods are used in robotics, target tracking, navigation, and large-scale data assimilation, where both accurate state estimates and calibrated uncertainty matter. A central challenge is balancing computational cost and scalability against the ability to represent complex posterior distributions, especially in high-dimensional systems with sparse observations or changing dynamics.

Latest papers

CardsList

61 papers9 papers in the last four weeks, up 13% on the four weeks before. 0.1% of all new papers.

  1. The Constitutional Filter: Bayesian Estimation of Compliant Agents

    Dec 24, 2024Simon Kohaut, Felix Divo, Benedict Flade +3Bayesian FilteringNeuro-Symbolic Reasoning