In-Network Processing for Timely Federated Multi-Task Learning over Wireless Edges

Shih-Chun Lin

Project runs from 08/01/2021 to 07/31/2022
$175,001

This project will employ in-network computation at 5G wireless edges and enable data processing to occur in the middle of the transmissions between multi-agents and remote cloud servers. Thus, it forges effective convergence of communication, computing, and learning with regard to wireless link bandwidths, available computing power, and collected data’s statistical features. Also, this project will provide a privacy-preserving framework from decentralized data by exploiting in-network processing operations. The proposed framework can achieve sophisticated and heavy-loaded machine learning algorithms through multiple low-end control units. It, in turn, preserves the data privacy for massive mobile user information in ICT or sensing information and intelligent manufacture/control commands in industrial scenarios. On the other hand, the proposed solution can also offload computation to the networking infrastructure, releasing the burden of multi-agents.