Next-Generation Communication Design for Distributed Intelligence over Wireless Networks

Wireless connectivity will always be the performance bottleneck for any distributed system that relies on information exchange among the network nodes to achieve a global objective.

Facing the rapid growth of decentralized Artificial Intelligence (AI) applications, the role of wireless communications needs to be re-defined rather than blindly following the traditional rate-driven design.This project aims at providing theoretical guidelines on communication design and signal processing for decentralized machine learning over wireless networks, with Federated Learning (FL) as the most representative example.

The main rationale behind this project is that the principle of information processing and transmission in a distributed learning system should adapt to the collective goal of the entire system.

New solutions will be proposed to alleviate the communication bottleneck in both server-based (centralized) and server-free (decentralized) FL systems, which are particularly important for supporting large-scale interconnected intelligent systems through next-generation wireless technology.This project will be carried out at Linköping University, with Zheng Chen as the project leader and primary investigator (PI).

The research activities will be carried out by Zheng Chen and one doctoral student.

The success of this project will unfold a wide range of promising research directions lying at the intersection of communication theory, machine learning, and distributed systems.