Our objective is to asses the performance of covariance matrix regularization methods on real world data, to provide points of reference for future applications. We analyse the following estimators: OAS, Rao-Blackwell-Ledoit-Wolf, Ledoit-Wolf in two versions, and Thresholding on data from several publicly available datasets (K9, Isolet, Slice, Gistette, S1 ADL1). We investigate through several norms the error of estimation from reduced data.
CITATION STYLE
Głomb, P., & Cholewa, M. (2015). Experimental evaluation of selected approaches to covariance matrix regularization. In Lecture Notes in Artificial Intelligence (Subseries of Lecture Notes in Computer Science) (Vol. 9120, pp. 391–401). Springer Verlag. https://doi.org/10.1007/978-3-319-19369-4_35
Mendeley helps you to discover research relevant for your work.