Delayed Gradient Averaging: Tolerate the Communication Latency in Federated Learning

Federated Learning is an emerging direction in distributed machine learning that enables jointly training a model without sharing the data.