A new intelligent pattern classifier based on structured sparse representation. (June 2020)

• Record Type: Journal Article
• Authors: Shen, Zhenyi; Man, Zhihong; Cao, Zhenwei; Zheng, Jinchuan
• Is Part Of: Computers & electrical engineering. Volume 84(2020)
• Publication Date: 2020-06
• Subjects: Intelligent classifier -- Pattern classification -- Sparse representation classifier -- Random nonlinear mapping -- Bayesian inference; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2020.106641
• Languages: English
• ISSNs: 0045-7906

Abstract

Abstract: A new intelligent pattern classifier based on structured sparse representation is developed, aiming to utilise the subspace structure to simplify the training process and reduce the required training sample size. It is seen that no mathematical optimization is involved in the training stage. In the developed classifier, the classification is made by selecting the optimal linear subspace that can represent data with the highest probability. A nonlinear mapping is designed to ensure data vectors in the feature space are more likely to live in their own subspaces. The sparse representation vectors of input data can then be computed in terms of a dictionary constructed by membership vectors of these subspaces. After that, the Bayesian inference based decision-making algorithm is implemented to use the statistical information imbedded in sparse representation vectors to classify the data efficiently. Lastly, a self-learning mechanism is developed to adjust the membership vectors of the dictionary to ensure the classifier's response to the subsequent application of similar input patterns will be enhanced. The experiment results show excellent classification performance and strong robustness of proposed intelligent pattern classifier.