美国伊利诺伊州理工学院程雨教授特邀讲座
主题： Machine Learning Meets Wireless Network Optimization
时间：2018年7月12日（星期四）上午10：30
地点：计算机与网络安全学院8A307
主办单位：计算机与网络安全学院、无线智能网络实验室
报告人简介：
Yu Cheng received B.E. and M.E. degrees in electronic engineering from Tsinghua University in 1995 and 1998, respectively, and a Ph.D. degree in electrical and computer engineering from the University of Waterloo, Canada, in 2003. He is now a full professor in the Department of Electrical and Computer Engineering, Illinois Institute of Technology. His research interests include wireless network performance analysis, network security, big data, cloud computing, and machine learning. He received a Best Paper Award at QShine 2007, IEEE ICC 2011, and a Runner-Up Best Paper Award at ACM MobiHoc 2014. He received the National Science Foundation (NSF) CAREER Award in 2011 and IIT Sigma Xi Research Award in the junior faculty division in 2013. He has served as Symposium Co-Chairs for IEEE ICC and IEEE GLOBECOM, and Technical Program Committee (TPC) CoChair for WASA 2011 and ICNC 2015. He is a founding Vice Chair of the IEEE ComSoc Technical Subcommittee on Green Communications and Computing. He was an IEEE ComSoc distinguished lecturer in 2016-2017. He is an Associate Editor for IEEE Transactions on Vehicular Technology and an IEEE senior member.
报告内容摘要：
With the rapid progress in the neural network based machine learning techniques, problems long thought to be hard like high-accuracy pattern recognition can now be solved with a reasonable amount of computing resources and data examples. Such techniques could find wide application in the area of wireless networking, characterized by abundant data and hard decision problems. In this talk, we present a pioneering study on using neural networks to assist in the classic network optimization task of energy-efficient flow planning, with various constraints including traffic demands, system throughput and link capacity. This is a well-known NP-hard problem and has stimulated a significant amount of effort in developing approximation algorithms. Taking a different approach, we exploit the neural network's pattern matching capabilities to prune the links unlikely to be used. This leads to a smaller-sized problem instance while preserving the optimality to the original problem. To this end, we design a learning framework to find the latent relationship between flow information and link usage by learning from past computation experience. Numerical results demonstrate that the proposed method could achieve the stated goal with a significantly reduced computation cost.
