QSplitFL: Capability Aware Deep Q-Learning for Optimal Split Point Selection in Split Federated Learning
Researchers have developed QSplitFL, a new framework using Deep Q-Learning to optimize split points in federated learning. This approach considers client hardware capabilities like CPU usage and memory, unlike previous methods that focused on model weights. QSplitFL aims to improve convergence speed and accuracy in federated learning scenarios with diverse devices, as demonstrated through experiments on various datasets and model architectures. AI
IMPACT Introduces a novel method for optimizing federated learning, potentially improving efficiency and accuracy on heterogeneous devices.