Momentum-Guided Federated Split Distillation for Edge Intelligence
September 28, 2026
The TeRR-SAtt and AMGF framework optimizes temporal edge intelligence by combining momentum-guided fusion with a temporal reservoir student attention design. On smart-building data, it reduces edge training latency by 65.5% and inference latency by 44.7% while improving local learning RMSE by 35.31%.
HOW THIS AFFECTS YOU
●
builderYou can deploy more efficient, personalized models to edge devices with significantly lower memory and CPU overhead.