Title: Threshold neural network and adaptive signal processing applications
Time: 10:30 on July 30, 2020
Tencent Conference ID: 445 500 884
Speaker: Professor Xiaoping Zhang
Moderator: Associate Professor Hu Menghan
Introduction of the speaker:
Professor Xiaoping Zhang is an academician of the Canadian Academy of Engineering. He received his bachelor and doctorate degrees from the Department of Electronic Engineering of Tsinghua University in 1992 and 1996, respectively. He received an MBA in Finance and Economics from the Booth School of Business at the University of Chicago and was honored as an outstanding graduate.
Academician Zhang Xiaoping is currently a full professor (tenured faculty) in the Department of Electrical, Computer and Biological Engineering, the director of the Communication and Signal Processing and Application Laboratory, and a professor of Finance in the Ted Rogers School of Management in Ryerson University, Canada. He once served as a graduate student and research supervisor of the department. In 2015 and 2017, Academician Zhang Xiaoping served as a visiting scientist in the Electronics Laboratory of Massachusetts Institute of Technology. He founded the financial big data search and analysis engine company EidoSerach and is the current CEO. Academician Zhang Xiaoping is committed to the research and development of signal processing and big data theory and application, mainly engaged in the research and development of statistical models, signal processing, machine learning and artificial intelligence, Internet of Things and electronic information systems, biological information and financial economic models, and big data. Academician Zhang Xiaoping is an internationally renowned expert in his research field and has worked in Wall Street and Silicon Valley industries. He was a visiting scientist at MIT and Harvard University. He has published more than 200 academic papers in top international journals and conferences and has a number of US patents, most of which have been converted into commercial products. Professor Zhang Xiaoping is currently the Senior Area Editor of IEEE Signal Processing Transactions and IEEE Image Processing Transactions. He has served as the Associate Editor of internationally renowned academic journals such as IEEE Signal Processing Transactions, IEEE Multimedia Processing Transactions, IEEE Image Processing Transactions, “IEEE Circuits and System Video Technology Transactions”, “IEEE Signal Processing Letters”, etc. He is currently the Vice Chairman of the Image, Video and Multidimensional Signal Processing Technology Committee of the IEEE Signal Processing Society, and the General Co-chair of the IEEE ICASSP 2021 General Assembly, the largest annual flagship of international signal processing. He also served as the Chairman of the 2017 and 2019 IEEE Global Conference on Signal and Information Processing (GlobalSIP) Financial and Business Information Processing Conference, and the Chairman of the 2015 IEEE Multimedia Signal Processing Annual Conference (MMSP2015). He is currently a member of the Steering Committee of the IEEE International Multimedia Conference (ICME). Professor Zhang has been invited to give tutorials report in various well-known international conferences such as ACM Multimedia Annual ACMMM2011, IEEE Circuits and Systems will ISCAS2013 and ISCAS2019, IEEE IMAGE PROCESSING will ICIP2013, IEEE Signal Processing will ICASSP2014, the International Neural Network Joint Assembly IJCNN2017. In 2019, he was selected for the IEEE Signal Processing Society Distinguished Lecture scholar for a term from January 2020 to December 2021. In 2020, he won the highest academic research award of Ryerson University, Sarwan Sahota Ryerson Outstanding Scholar Award.
In this talk, a system framework of nonlinear thresholding for adaptive signal processing, namely thresholding neural network (TNN), is presented. Several types of thresholding functions are created to serve as activation functions. Unlike the standard thresholding functions, the new thresholding functions are infinitely differentiable. By using the new thresholding functions, some gradient-based learning algorithms become possible or more effective. General optimal performances of TNNs are analyzed. Gradient-based adaptive learning algorithms are presented to seek the optimal solution for noise reduction. The algorithms include supervised and unsupervised batch learning as well as supervised and unsupervised stochastic learning. It is indicated that the TNN with the stochastic learning algorithms can be used as a novel nonlinear adaptive filter. Numerical results show that the TNN is very effective in finding the optimal solutions of thresholding methods in an MSE sense and usually ou tperforms other noise reduction methods. Especially, it is shown that the TNN based nonlinear adaptive filtering outperforms the conventional linear adaptive filtering in both optimal solution and learning performance.