ASFL: An Adaptive Model Splitting and Resource Allocation Framework for Split Federated Learning

Researchers propose ASFL, an adaptive split federated learning framework that optimizes machine learning model training across wireless networks by splitting computation between clients and central servers. The framework reduces training delay by up to 75% and energy consumption by 80% compared to baseline approaches while maintaining faster convergence rates.

• →ASFL enables federated learning with adaptive model splitting between clients and central servers to optimize resource allocation.
• →The framework addresses limited computation resources of edge clients through intelligent workload distribution.
• →An online optimization enhanced block coordinate descent algorithm solves the joint optimization problem iteratively.
• →Experimental results show up to 75% reduction in delay and 80% reduction in energy consumption compared to baseline schemes.
• →The approach maintains data privacy while improving training efficiency through adaptive resource management.