On kernel design for regularized LTI system identification
For researchers in system identification, this work provides principled kernel design methods, but the results are theoretical without concrete performance numbers.
This paper addresses kernel design for regularized LTI system identification, proposing two methods from machine learning and system theory perspectives. The analysis enhances understanding of existing kernels and guides design of new ones.
There are two key issues for the kernel-based regularization method: one is how to design a suitable kernel to embed in the kernel the prior knowledge of the LTI system to be identified, and the other one is how to tune the kernel such that the resulting regularized impulse response estimator can achieve a good bias-variance tradeoff. In this paper, we focus on the issue of kernel design. Depending on the type of the prior knowledge, we propose two methods to design kernels: one is from a machine learning perspective and the other one is from a system theory perspective. We also provide analysis results for both methods, which not only enhances our understanding for the existing kernels but also directs the design of new kernels.