In this paper a method of faster training of the ensembles of the tensor classifiers based on the Higher-Order Singular Value Decomposition is presented. The method relies on the fixed-point method of eigenvector computation which is employed at the stage of subspace construction of the flattened versions of the input tensor patterns. As verified experimentally, the proposed method allows up to five times speed-up factor at no significant difference in accuracy.
CITATION STYLE
Cyganek, B., & Woźniak, M. (2016). On robust computation of tensor classifiers based on the higher-order singular value decomposition. In Advances in Intelligent Systems and Computing (Vol. 465, pp. 193–201). Springer Verlag. https://doi.org/10.1007/978-3-319-33622-0_18
Mendeley helps you to discover research relevant for your work.