FedDQ: Communication-Efficient Federated Learning with Descending Quantization | IEEE Conference Publication | IEEE Xplore