Xuelong Li is a full professor at The Xi'an Institute of Optics and Precision Mechanics of Chinese Academy of Sciences, and an associate director of The State Key Laboratory of Transient Optics and Photonics. He founded The Center for OPTical IMagery Analysis and Learning (OPTIMAL). He received his B. Eng and Ph. D. degrees from the University of Science and Technology of China (USTC).

His research focuses on visual information processing and searching; video surveillance; and multimedia communication. He has 300+ research papers, including 120+ in IEEE transactions and 70+ in Elsevier/Springer journals, with 5000+ citations and a few paper awards (e.g., Best/Hot Papers). He was/is an editor of 19 journals, including Pattern Recognition (Elsevier) and six IEEE transactions. He is a Member of Global Young Academy, an Academician of the IEAS (International Eurasian Academy of Sciences), and a Fellow of the IEEE, IEE/IET, SPIE, OSA, IAPR, BCS, and HEA.

Matrix Decomposition for Big Visual Data Analysis

Abstract:

Conventional learning models process data represented by vectors, although visual data are naturally of tensor format. Tensor data analysis and its applications to optical imagery analysis and learning are emergent yet critical research areas, which have witnessed a rapid proliferation. Tensor learning as a new research direction is in the making, which opens a different perspective for understanding structural data. In this area, the most critical and fundamental research tasks are to construct new data representations, study algorithm convergence and stability, and explore the statistical theory to understand the learning mechanism. Professor Xuelong Li proposed a generalized tensor learning framework, through which he solved the problem of effectively training, supervised classification, distance metric learning, manifold learning, and comprehensively considering the structural dependency relationship encoded in data. The application of tensor learning to optical imagery data effectively improves the quality of imagery data and changes the way to understand the data. This has been an essential tool to resolve important technical difficulties, and has been widely recognized yet highly evaluated by international peers. Tensor learning has strong potential to back the development of machine learning, computer vision, multimedia, pattern recognition and data mining.