Prof. Nikolaos M. Freris

IEEE Senior Member

School of Computer Science

University of Science and Technology of China, China

Speech Title:  

Communication-efficient distributed machine learning for AIoT

Abstract:

The ever-emerging paradigm of AIoT targets solutions that capitalize on the computational power of Internet-of-Things (IoT) devices to enable Artificial Intelligence (AI) in system operations. This is accomplished by shifting away from centralized cloud computing to decentralized edge computing. To this end, distributed optimization is the workhorse tool for machine learning in a multi-agent network. Despite the large body of work on the subject, it is imperative to devise new methods tailored to address the distinctive challenges of AIoT systems such as: a) huge-scale, b) stringent constraints on communication and latency, and c) heterogeneity in terms of device computational capabilities as well as data distributions. This keynote seminar will present a novel framework for distributed optimization based on operator splitting that specifically targets communication efficiency and adaptability to user heterogeneity. The product is a range of asynchronous algorithms (first and second order) for both convex and non-convex problems with rigorous analysis in terms of convergence rate and communication costs. In particular, I will present a new method for federated learning (distributed deep learning in large heterogeneous networks) with state-of-the-art performance in terms of convergence speed, communication savings, and accuracy.

Bio:

Nick Freris is Professor in the School of Computer Science at USTC, and Vice Dean of the International College. He received the Diploma in Electrical and Computer Engineering from the National Technical University of Athens (NTUA), Greece, in 2005, and the M.S. degree in Electrical and Computer Engineering, the M.S. degree in Mathematics, and the Ph.D. degree in Electrical and Computer Engineering all from the University of Illinois at Urbana-Champaign (UIUC) in 2007, 2008, and 2010, respectively. His research lies in AIoT/CPS/IoT: machine learning, distributed optimization, data mining, wireless networks, control, and signal processing, with applications in power systems, sensor networks, transportation, cyber security, and robotics. Dr. Freris has published several papers in high-profile conferences and journals held by IEEE, ACM, and SIAM and holds three patents. His research has been sponsored by the Ministry of Science and Technology of China, Anhui Dept. of Science and Technology, Tencent, and NSF, and was recognized with the National High-level Talent award, the USTC Alumni Foundation Innovation Scholar award, and the IBM High Value Patent award. Previously, he was with the faculty of NYU and, before that, he held

senior researcher and postdoctoral researcher positions at EPFL and IBM Research, respectively. Dr. Freris is a Senior Member of ACM and IEEE, and a member of CCF and SIAM.