EE Colloquium: On federated learning over wireless fading channels

Abstract: Federated learning (FL) is an emerging distributed machine learning (ML) paradigm that enables many clients to collaboratively train a ML model under the coordination of a central server. This paradigm naturally fits into the cellular architecture where the communication phases take place over wireless links, which represents a formidable challenge for current wireless systems because FL has unique characteristics that are fundamentally different from many of the use cases in today's wireless systems. 

In this talk, we will discuss two design examples of federated learning over wireless communications, to highlight the benefits of tight integration of ML and communications. In the first part, we will show that deep fading effect, the well-known typical error event for communication over wireless channels, is particularly detrimental for FL. In fact, the celebrated FedAvg and its several variants break down for FL tasks when deep fading exists in the communication phase. We then propose an optimal global model aggregation method at the parameter server, which allocates different weights to different clients based not only on their learning characteristics but also on the instantaneous channel conditions, hence taking into consideration the heterogeneous and highly dynamic channel fading effect across clients. In the second part, we will discuss how massive multiple-input and multiple-output (MIMO) can benefit FL. A novel uplink communication method, coined random orthogonalization, is proposed based on the tight coupling of FL model aggregation and two unique characteristics of massive MIMO – channel hardening and favorable propagation. As a result, random orthogonalization can achieve natural over-the-air model aggregation without requiring transmitter side channel state information, while significantly reducing the channel estimation overhead at the receiver. We will also discuss potential extensions of random orthogonalization if time permits. 

Bio: Dr. Cong Shen is an Assistant Professor of the Electrical and Computer Engineering Department at University of Virginia (UVa). He received his B.S. and M.S. degrees from Tsinghua University, and Ph.D. from University of California Los Angeles (UCLA), all in electrical engineering. He has extensive industry experience, having worked for Qualcomm Research, SpiderCloud Wireless, Silvus Technologies, and, in various full time and consulting roles. His general research interests lie in wireless communications and machine learning, with the current focus on multi-armed bandits, reinforcement learning, federated learning, and their engineering applications. He was the recipient of the Best Paper Award in 2021 IEEE International Conference on Communications (ICC). Currently, he serves as an editor for the IEEE Transactions on Green Communications and Networking and the IEEE Wireless Communications Letters. He was the TPC co-chair of the Wireless Communications Symposium of IEEE Globecom 2021 and actively serves as (senior) program committee members/reviewers for NeurIPS, ICML, ICLR, AISTATS, IJCAI, and AAAI. 


Share this event

facebook linked in twitter email

Media Contact: Iam-Choon Khoo



The School of Electrical Engineering and Computer Science was created in the spring of 2015 to allow greater access to courses offered by both departments for undergraduate and graduate students in exciting collaborative research fields.

We offer B.S. degrees in electrical engineering, computer science, computer engineering and data science and graduate degrees (master's degrees and Ph.D.'s) in electrical engineering and computer science and engineering. EECS focuses on the convergence of technologies and disciplines to meet today’s industrial demands.

School of Electrical Engineering and Computer Science

The Pennsylvania State University

207 Electrical Engineering West

University Park, PA 16802


Department of Computer Science and Engineering


Department of Electrical Engineering