A nonlinear tensor-based machine learning algorithm for image classification

In recent year, the tensor theory has been frequently incorporated to machine learning, because of the various advantages of tensor-based machine learning over vector-based machining learning: the ability to preserve the spatiotemporal information, allowing full utilization of the data, and the suitability for solving high-dimensional problems with a small sample size. Considering the suitability of tensor algorithm for classical high-dimensional, small-sample problems, this paper probes into the nonlinear classification problem with tensor representation, and designs a tensor-based nonlinear classification algorithm, namely, the kernel-based STM (KSTM). The maximum margin principle was adopted for the classification by the KSTM: the two types of samples are separated by the decision hyperplane as far as possible in the tensor space. Through numerical experiments, it is proved that the KSTM achieved better classification accuracy than the linear method, especially for the high-dimensional problem with a small sample size. © 2019 Lavoisier. All rights reserved.