Gray-Box Identification for High-Dimensional Manifold Constrained Regression
2009 (English)Report (Other academic)
High-dimensional gray-box identification is a fairly unexplored part of system identification. Nevertheless, system identification problems tend to be more high-dimensional nowadays. In this paper we deal with high-dimensional regression with regressors constrained to some manifold. A recent technique in this class is weight determination by manifold regularization (WDMR). WDMR, however, is a black-box identification method. We show how WDMR can be extended to a gray-box method and illustrate the scheme with some examples.
Place, publisher, year, edition, pages
Linköping: Linköping University Electronic Press, 2009. , 8 p.
LiTH-ISY-R, ISSN 1400-3902 ; 2896
Nonparametric identification, system identification, Regularization, Non-parametric regression, identification algorithms
IdentifiersURN: urn:nbn:se:liu:diva-56203ISRN: LiTH-ISY-R-2896OAI: oai:DiVA.org:liu-56203DiVA: diva2:317006