Adaptive Joint Compression and Synchronisation in Federated Split Learning for IoT Rainfall Prediction
arXiv:2606.25003v1 Announce Type: new Abstract: Federated split learning (FSL) enables collaborative training across bandwidth-constrained IoT devices, but repeated activation and gradient exchange creates a communication bot-tleneck. Prior work optimises either activation compression or synchronisation frequency in isolation. This paper presents an FSL framework for IoT rainfall prediction that jointly regulates activation compression and the synchronisation interval \r{ho} via a latency driven scheduler on a server with per client EMA smoothing. The system is evaluated on hourly ERA5 data fr
The rapid deployment of IoT devices and increasing demand for real-time analytics in bandwidth-constrained environments necessitates more efficient federated learning techniques.
This development allows for more robust and resource-efficient AI model training directly on edge devices, which is critical for scalable, privacy-preserving, and low-latency applications.
The ability to jointly optimize communication bottlenecks in federated split learning makes advanced AI applications more viable for widespread deployment in distributed IoT networks.
- · IoT device manufacturers
- · Edge AI providers
- · Smart city developers
- · Climate data analytics
- · Centralized cloud AI services for IoT
- · High-bandwidth network infrastructure dependent solutions
More accurate and localized rainfall predictions enabled by distributed AI on IoT at reduced operational costs.
Reduced infrastructure demands for IoT deployments, accelerating adoption in remote or resource-limited regions.
Enhanced data privacy and security as raw data remains local, fostering greater trust and wider application of AI in sensitive domains.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG