Repairability of Inexact Solvers in Recursive State Estimation with Machine Learning
A certified repair framework evaluates inexact and quantum linear solvers inside Kalman estimation, including a power-grid study.
The paper analyzes when a bounded subspace correction can satisfy a local admissibility tolerance inside a linear Kalman estimator and how solver defects affect finite-horizon covariance. It separates current solve error from inherited gain drift and identifies opposing quartic contributions that can produce quadratic under- or overprediction. Learned corrections remain governed by a learner-independent residual certificate and fallback. In a power-grid study, learned correction reduced the conjugate-gradient iterations required for certified deployment, and variational and annealing quantum solvers were executed through the same interface.
- Bounded corrections are certified against a local Kalman admissibility tolerance.
- Learned fixes reduced conjugate-gradient iterations in a power-grid study.
- Quantum variational and annealing solvers used the same certified interface.
- Opposing quartic terms can cause quadratic covariance under- or overprediction.
From the original article
1 min read at arxiv.orgRecursive state estimation often executes approximate numerical solutions inside a feedback loop, where highly accurate local steps do not guarantee better overall results. For a fixed linear Kalman model, we characterize when a correction within a prescribed subspace and norm budget can meet a local admissibility tolerance, and how the defects actually executed affect the finite-horizon covariance response.
By Yanjun Ji · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.