Least squares twin support tensor machine for classification

Least Squares Support Vector Machine (LS-SVM) and Twin Support Vector Machine (TSVM) are effective learning methods for classification based on Support Vector Machine which has been widely used in many aspects and received extensive attention by academic community. At present, data representation is one of the core problems in machine learning. In practice, many objects are naturally represented by tensors. In this paper, we propose Least Squares Twin Support Tensor Machine (LS-TSTM) which based on tensor data. We use two non-parallel hyperplanes and least squares idea for classification, which is different form Support Tensor Machine (STM). LS-TSTM combines the characteristics of many learning machines. It makes full use of the structural information of the data, and has the features of less computation cost and higher precision. The numerical experiments of the two-class classification problem for tensor data show that LS-TSTM has its advantages compared with other learning machines.