Work

An Efficient Federated Learning with Correlation-Based Pruning: Improving Accuracy under Layer-Wise Differential Privacy