Federated Edge Learning is the application of federated learning specifically to edge devices at the network edge — combining FL with mobile edge computing (MEC) to enable collaborative model training across edge nodes while leveraging edge computing infrastructure for efficient aggregation.

Federated Edge Architecture

• Edge Devices: Sensors, equipment controllers, and IoT devices perform local model training.
• Edge Server: Local aggregation at the edge server (within the fab or site) — reduces latency and bandwidth.
• Cloud: Optional global aggregation across sites — hierarchical FL architecture.
• Over-the-Air: Wireless aggregation (analog over-the-air computation) for ultra-efficient communication.

Why It Matters

• Low Latency: Edge aggregation is faster than cloud aggregation — critical for time-sensitive applications.
• Bandwidth: Aggregating at the edge reduces WAN bandwidth requirements.
• Semiconductor: Edge devices in a fab can federate locally for real-time process optimization.

Federated Edge Learning is collaborative learning at the edge — combining federated learning with edge computing for efficient, low-latency model training.