讲座题目(Title of Lecture)：Deep Learning Foundations and Applications in Smart Grid

讲座时间(Time of Lecture)：2020年12月28日星期一10：00-11：00

讲座地点(Site of Lecture)：腾讯会议ID：132 728 925

主讲人(Lecturer)：Dr. Bo Tang, Mississippi State University

报告人简介(Introduction of Lecturer)：

Dr. Bo Tang (唐波) is currently an Assistant Professor in the Department of Electrical and Computer Engineering at Mississippi State University. He received his B.S. and M.S. degrees from Central South University in 2007, and Institute of Electronics, Chinese Academy of Sciences in 2010, respectively. He received the Ph.D. degree in electrical and computer engineering from University of Rhode Island (Kingstown, RI). His research interests lie in the general areas of Statistical Machine Learning, Deep Learning, Cyber Security, and their applications in Robotics and Cyber Physical Systems. In particular, he is interested in developing new fundamental machine learning algorithms (e.g., Bayesian learning, transfer learning and deep learning) and building complex embedded systems (e.g., energy systems, robotics and autonomous vehicles) that are robust, adaptive and fault tolerant to uncertain environments. 

Dr. Tang is also the recipient of NIJ New Investigator/Early Career Award (2018), Chinese Government Award for Outstanding Students Abroad (2016), Best Paper Award in IEEE CCWC (2018), Best Student Paper Award in IJCNN (2016), Junior Faculty Travel Award by Army Research Office (2016), Travel Award by IEEE CNAS (2015), and IEEE Computational Intelligence Magazine Publication Spotlight Paper (2015).

讲座内容(Content of Lecture)：

Tremendous success of deep learning has been recently demonstrated in a variety of complex and challenging application domains, such as natural language processing, computer vision, and robotics. In this two-part talk, I will first focus on the introduction of a new type of neural architectures with memory, also called memory neural networks, which couple a neural network with an external memory matrix. The use of explicit memory allows the intelligent system explore deeper and longer-term patterns from the data. Then I will present our efforts of developing a new differential neural computer (DNC) with a convertible short-term and long-term memory (CSLM) mechanism. This is mainly motivated by human brains where both short-term and long-term memory exist and they can be converted over time for effective and efficient information encoding, storage, and retrieval. Our experimental results on challenging Question-Answering and Copy/Repeat Copy tasks have demonstrated its state-of-the-art performance. In the second part of this talk, I will discuss how machine learning, particularly deep learning, can facilitate the field of power system by improving the security and efficiency. Two examples of our current works in intrusion detection and power system simulation will be presented and discussed.