Title: Network Latency Prediction for Personal Devices: from Matrix Completion to Tensor Approximation

Abstract:

With the popularity of real-time apps like live chat and gaming, Internet latency prediction between personal devices becomes increasingly important. Traditional approaches recover all-pair latencies in a network from sampled measurements using either Euclidean embedding or matrix factorization. However, these schemes become less effective to estimate latencies between personal devices, due to unstable network conditions, triangle inequality violation in the network and unknown rank of latency matrices. In this talk, I will describe new methods for both static latency estimation as well as dynamic estimation given 3D latency measurements sampled over time. We have proposed a new class of low-rank matrix completion algorithms by iteratively minimizing a carefully weighted Schatten-p norm. We further extend our algorithms to dynamic latency estimation via tensor approximation. Performance evaluations driven by real-world measurements show that our approaches significantly outperform various state-of-the-art latency prediction techniques.


Short bio:

Di Niu received the B.Engr. degree from the Department of Electronics and Communications Engineering, Sun Yat-sen University, China, in 2005 and the M.A.Sc. and Ph.D. degrees from the Department of Electrical and Computer Engineering, University of Toronto, in 2009 and 2013. Since September, 2012, he has been with the Department of Electrical and Computer Engineering at the University of Alberta, where he is currently an Assistant Professor. His research interests include cloud computing and storage, computer networking, data mining and statistical machine learning for social economic computing, distributed optimization, and network coding.