Tensor Learning Team Seminar (Talk by Dr. Xile Zhao, University of Electronic Science and Technology of China).

Title: High-Dimensional Image Recovery: When Matrix/Tensor Decomposition Meets Unsupervised Learning

Speaker: Dr. Xile Zhao, University of Electronic Science and Technology of China

Abstract:
Recently, low-rank matrix/tensor decomposition methods have received increasing attention for high-dimensional data recovery. However, only considering the low-rank structure of high-dimensional data is not sufficient for high-dimensional data recovery, especially for extremely complex imaging scenarios. In this talk, we will discuss how to bring into play the respective strengths of matrix/tensor decomposition and unsupervised learning for high-dimensional data recovery. Extensive numerical examples including inpainting, denoising, and snapshot compressed sensing are delivered to demonstrate the superiority of the suggested methods over state-of-the-art methods.

Bio: Xi-Le Zhao is currently a professor with the School of Mathematical Sciences, University of Electronic Science and Technology of China (UESTC). He received the Ph.D. degree from UESTC. He worked as a post-doc with Prof. Michael K. Ng at Hong Kong Baptist University from 2013 to 2014. He worked as a visiting scholar with Prof. José M. Bioucas Dias at University of Lisbon from 2016 to 2017. His research interests include models, algorithms, and theories for the low-level inverse problems of multi-dimensional images. He has published two book chapters and more than 50 academic papers in top journals such as SIAM J. Imaging Sci., SIAM J. Scientific Comput., IEEE Trans. Image Process., IEEE Trans. Neural Netw. Learn. Syst., IEEE Trans. Cybernetics, IEEE Trans. Geosci. Remote Sens., IEEE Trans. Comput. Imaging, CVPR, and AAAI. He has received many awards such as the First Prize of Science and Technology Progress Award of Sichuan Province and the Second Prize in the 6th Contest of Outstanding Young Essay awarded by the Chinese Society of Computational Mathematics.