Retrospective Sensing Based on Federated Learning in the IoT

Rafiq Mazen Kamel, Amr El Mougy

Abstract: Knowledge in the IoT is typically distributed and sparse. Combining information from multiple sources remains a challenge to researchers. In addition, due to the strict capabilities of sensing devices, they may go to sleep for extended durations and miss critical data. In this paper, we propose a system that is capable of retrospective sensing based on federated machine learning models. The system is implemented on a distributed edge computing architecture and is capable of fusing learned data from several sensors to produce accurate estimations of missed data. The proposed system is evaluated using computer simulations on real and synthetic datasets and shows high accuracy in predicting missed data.