Abstract: Massive networks of wireless devices offer new opportunities for large-scale machine learning, but their decentralized nature and resource-limited wireless connectivity pose fundamental challenges. This talk will discuss fully decentralized learning systems in which devices collaboratively train a model without relying on a central coordinator.

The talk will focus on decentralized gradient descent, whose standard implementation requires consensus over a communication graph. While effective in principle, its naive wireless implementation requires topology knowledge, channel-aware control, and interference management, making it difficult to scale efficiently. The talk will then present a non-coherent over-the-air consensus scheme that exploits, rather than avoids, simultaneous wireless transmissions. By using energy superposition, the method enables one-shot consensus without channel state information or explicit knowledge of the graph topology.

Recent extensions of this approach will also be discussed, including robustness to external interference and operation over broader classes of channel models beyond Rayleigh flat fading. The talk will present the core algorithmic ideas, convergence guarantees, and numerical results for decentralized learning over wireless networks.

Bio: Nicolò Michelusi is an Associate Professor in the School of Electrical, Computer and Energy Engineering at Arizona State University. He received the B.Sc., M.Sc., and Ph.D. degrees from the University of Padova, Italy, and the M.Sc. degree in Telecommunications Engineering from the Technical University of Denmark as part of the T.I.M.E. double degree program. Before joining ASU, he was a Postdoctoral Research Fellow at the University of Southern California and an Assistant Professor

at Purdue University.

His research interests include wireless communications and networks, machine learning over wireless systems, decentralized and federated learning, and stochastic optimization. Dr. Michelusi is a Senior Member of the IEEE. He has served as Associate Editor for the IEEE Transactions on Wireless Communications, Editor for the IEEE Transactions on Communications, and Guest Editor for a special issue of the IEEE Journal on Selected Areas in Communications on distributed learning over wireless networks. He received the NSF CAREER Award in 2021, the IEEE Communication Theory Technical Committee Early Achievement Award in 2022, the IEEE Communications Society William R. Bennett Prize in 2024, and an IEEE ICC 2025 Best Paper Award for his work on non-coherent over-the-air federated learning.