学术报告 | AI in Networks: From Intelligent Traffic Engineering to Intelligent RF Sensing

发布者:孙威发布时间:2019-07-02浏览次数:13

报告题目:AI in Networks: From Intelligent Traffic Engineering to Intelligent RF Sensing
报告人:Dr. Pu Wang, Assistant Professor, University of North Carolina at Charlotte, USA
时  间:2019年7月5号上午10:00am-11:00am
地  点:无线谷A1号楼1306会议室


报告摘要:
    Artificial Intelligence (AI) and Machine Learning (ML) approaches have emerged in the networking domain with great promise. In this talk, I will present our recent progresses in such emerging field of AI in Networks. First, my talk will focus on our research on exploiting multi-agent reinforcement learning (MA-RL) approaches for stochastic optimization in large-scale computer networks. The key contribution of our MA-RL approaches is to enable model-free optimization of end-to-end traffic engineering in computer networks, which is very difficult, if not impossible, to achieve using conventional stochastic optimization schemes. To enable such learning-based network optimization, we are also developing modular and composable network intelligence system based on SDN technologies. This system provides modules and their extensions that can be assembled in various combinations to generate specific MA-RL algorithms.
    The second part of my talk will focus on deep learning-based wireless sensing and its application in RF-biometric identification. In particular, RF biometrics are human’s unique physical and behavioral characteristics, e.g., gait & keystroke, which are extracted from RF signals of ubiquitous wireless systems (e.g.,WiFi). RF-biometrics identification has great advantages of long-range detection, independence of lighting conditions, ability to penetrate materials (e.g., clothing and wall), excellent performance under adverse environmental conditions (e.g., smoke, fire, rain, fog, dust and snow), and inherent privacy-preserving capability. Despite these advantages, exiting RF biometrics identification systems have limited identification accuracies (around 87%) and are limited to a small number of people (up to 10). To address such limitations, we are developing deep RF-sensing systems, which exploit deep neural networks to automatically extract the salient and representative features from rich and high-dimension wireless signatures in different frequency bands (e.g., sub-5GHz and mmWave bands). Our preliminary systems can achieve over 95% identification accuracy over a dataset of 50 people.


个人简历:
     Pu Wang received B.Eng degree in Electrical Engineering from Beijing Institute of Technology, China, in 2003, and M.Eng degree in Electrical and Computer Engineering from Memorial University of Newfoundland, Canada, in 2008. He received the Ph.D. degree in Electrical and Computer Engineering from the Georgia Institute of Technology, Atlanta, GA, in August 2013, under the guidance of Professor Ian F. Akyildiz. Currently, he is an Assistant Professor with the Department of Computer Science at the University of North Carolina at Charlotte. Prior to joining UNCC, he was an assistant professor with Department of Electrical Engineering and Computer Science at Wichita State University from 2013 to 2017. His current research interests focus on optimization and machine learning in networked systems, with applications in Software Defined Networking, Cyber-Physical Systems, Internet of Things, Cognitive Radio Networks, and Electromagnetic Nanonetworks