TY - JOUR
T1 - Regularized LTI System Identification with Multiple Regularization Matrix
AU - Chen, Tianshi
AU - Andersen, Martin S.
AU - Mu, Biqiang
AU - Yin, Feng
AU - Ljung, Lennart
AU - Qin, S. Joe
PY - 2018
Y1 - 2018
N2 - Regularization methods with regularization matrix in quadratic form have received increasing attention. For those methods, the design and tuning of the regularization matrix are two key issues that are closely related. For systems with complicated dynamics, it would be preferable that the designed regularization matrix can bring the hyper-parameter estimation problem certain structure such that a locally optimal solution can be found efficiently. An example of this idea is to use the so-called multiple kernel Chen et al. (2014) for kernel-based regularization methods. In this paper, we propose to use the multiple regularization matrix for the filter-based regularization. Interestingly, the marginal likelihood maximization with the multiple regularization matrix is also a difference of convex programming problem, and a locally optimal solution could be found with sequential convex optimization techniques.
AB - Regularization methods with regularization matrix in quadratic form have received increasing attention. For those methods, the design and tuning of the regularization matrix are two key issues that are closely related. For systems with complicated dynamics, it would be preferable that the designed regularization matrix can bring the hyper-parameter estimation problem certain structure such that a locally optimal solution can be found efficiently. An example of this idea is to use the so-called multiple kernel Chen et al. (2014) for kernel-based regularization methods. In this paper, we propose to use the multiple regularization matrix for the filter-based regularization. Interestingly, the marginal likelihood maximization with the multiple regularization matrix is also a difference of convex programming problem, and a locally optimal solution could be found with sequential convex optimization techniques.
KW - regularization methods
KW - sequential convex optimization
KW - System identification
UR - http://www.scopus.com/inward/record.url?scp=85054358651&partnerID=8YFLogxK
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-85054358651&origin=recordpage
U2 - 10.1016/j.ifacol.2018.09.121
DO - 10.1016/j.ifacol.2018.09.121
M3 - 21_Publication in refereed journal
VL - 51
SP - 180
EP - 185
JO - IFAC-PapersOnLine
JF - IFAC-PapersOnLine
SN - 2405-8963
IS - 15
ER -