Abstract: The adaptive gradient descent algorithm in federated learning (FL) determines the speed of model convergence and has been widely studied and applied in recent years. Although the adaptive ...