A Low-Rank Approximation-Based Transductive Support Tensor Machine for Semisupervised Classification

摘要: In the fields of machine learning, pattern recognition, image processing, and computer vision, data are usually represented by tensors. For semisupervised tensor classification, existing transductive support (TSTM) needs to resort iterative technique, which is very time-consuming. order overcome this shortcoming, in paper, we extend concave-convex procedure-based vector (CCCP-TSVM) patterns propose a low-rank approximation-based TSTM, rank-one decomposition used compute inner product Theoretically, TSTM (CCCP-TSTM) an extension linear CCCP-TSVM patterns. When input vectors, CCCP-TSTM degenerates into CCCP-TSVM. A set experiments conducted on 23 classification tasks, generated from seven second-order face sets, three third-order gait two illustrate performance CCCP-TSTM. The results show that compared with provides significant gain terms test accuracy training speed.