Embedding of the Extended Euclidean Distance into Pattern Recognition with Higher-Order Singular Value Decomposition of Prototype Tensors

摘要

The paper presents architecture and properties of the ensemble of the classifiers operating in the tensor orthogonal spaces obtained with the Higher-Order Singular Value Decomposition of prototype tensors. In this paper two modifications to this architecture are proposed. The first one consists in embedding of the Extended Euclidean Distance metric which accounts for the spatial relationship of pixels in the input images and allows robustness to small geometrical perturbations of the patterns. The second improvement consists in application of the weighted majority voting for combination of the responses of the classifiers in the ensemble. The experimental results show that the proposed improvements increase overall accuracy of the ensemble.