Work

Latency-Efficient Wireless Federated Learning With Sparsification and Quantization for Heterogeneous Devices