Interpretable Gradient Descent for Kalman Gain
Interpretable Gradient Descent for Kalman Gain
We derive a decomposition for the gradient of the innovation loss with respect to the filter gain in a linear time-invariant system, decomposing as a product of an observability Gramian and a term quantifying the ``non-orthogonality" between the estimation error and the innovation. We leverage this decomposition to give a convergence proof of gradient descent to the optimal Kalman gain, specifically identifying how recovery of the Kalman gain depends on a non-standard observability condition, and obtaining an interpretable geometric convergence rate.
M. A. Belabbas、A. Olshevsky
自动化基础理论
M. A. Belabbas,A. Olshevsky.Interpretable Gradient Descent for Kalman Gain[EB/OL].(2025-07-22)[2025-08-10].https://arxiv.org/abs/2507.14354.点此复制
评论