We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Systems and Control

Title: Error Bounds for Kernel-Based Linear System Identification with Unknown Hyperparameters

Abstract: The kernel-based method has been successfully applied in linear system identification using stable kernel designs. From a Gaussian process perspective, it automatically provides probabilistic error bounds for the identified models from the posterior covariance, which are useful in robust and stochastic control. However, the error bounds require knowledge of the true hyperparameters in the kernel design and are demonstrated to be inaccurate with estimated hyperparameters for lightly damped systems or in the presence of high noise. In this work, we provide reliable quantification of the estimation error when the hyperparameters are unknown. The bounds are obtained by first constructing a high-probability set for the true hyperparameters from the marginal likelihood function and then finding the worst-case posterior covariance within the set. The proposed bound is proven to contain the true model with a high probability and its validity is verified in numerical simulation.
Subjects: Systems and Control (eess.SY); Machine Learning (stat.ML)
Cite as: arXiv:2303.09842 [eess.SY]
  (or arXiv:2303.09842v1 [eess.SY] for this version)

Submission history

From: Mingzhou Yin [view email]
[v1] Fri, 17 Mar 2023 08:52:16 GMT (183kb)

Link back to: arXiv, form interface, contact.