The Score Kalman Filter

Kaito Iwasaki, Anthony Bloch, Taeyoung Lee, Maani Ghaffari · 2026

Read The Score Kalman Filter by Kaito Iwasaki, Anthony Bloch, Taeyoung Lee, Maani Ghaffari (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-16644.txt.

Description

Nonlinear Bayesian filtering struggles to represent belief distributions, and moment filters with polynomial moments need maximum-entropy closure whose partition function cost grows fast with dimension. Iwasaki's team avoids that function outright, using score matching and Stein's identity to reduce the density fit to one linear solve. Their Score Kalman Filter reportedly recovers the classical information-form Kalman filter as one special case. On coupled-oscillator networks up to size 20, it reportedly beats extended, unscented, ensemble, and particle baselines on error.

Catalog record

CREATORKaito Iwasaki, Anthony Bloch, Taeyoung Lee, Maani Ghaffari
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE167558
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.16644

Browse the full library · About the Tower of Babel