Research map / Probabilistic and scientific modeling / Dynamical systems

Diff-Kalman: Difference-Driven Learning for Structure-Preserving Kalman Filtering

NeurIPS 2026 · Probabilistic and scientific modeling › Dynamical systems · approach: Data assimilation from sparse observations

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes Diff-Kalman, a difference-driven learning approach for structure-preserving Kalman filtering that addresses posterior-only optimization and unstable end-to-end prediction of high-dimensional Kalman parameters.

Problem

State estimation for dynamic systems, where existing Kalman filters and data-driven methods optimize only the posterior without prior modeling and predict high-dimensional Kalman parameters end-to-end, leading to unstable training.

Approach

Uses difference-driven learning within a Kalman filter structure to model the prior and avoid end-to-end prediction of high-dimensional Kalman parameters, preserving the filter's structure; specific mechanism not stated in the abstract.

Experimental setup

Builds on: Kalman filter

Closest papers by meaning

• Predictively-Oriented Kalman Filtering (NeurIPS 2026 · similarity 0.87)
• Implicit Maximum a Posteriori Filtering via Adaptive Optimization (ICLR 2024 · similarity 0.85)
• Gated Inference Network: Inference and Learning State-Space Models (NeurIPS 2024 · similarity 0.85)
• Markovian Dynamics Enforcer: Feasibility Preserving Correction on Learned Dyanmics Manifolds (NeurIPS 2026 · similarity 0.84)
• Nonlinear Filtering with Brenier Optimal Transport Maps (ICML 2024 · similarity 0.84)
• Outlier-robust Kalman Filtering through Generalised Bayes (ICML 2024 · similarity 0.83)