Delayed Random Partial Gradient Averaging for Federated Learning
This addresses scalability issues for federated learning systems with limited bandwidth and high latency, though it appears incremental as it builds on existing gradient averaging methods.
The paper tackles communication bottlenecks in federated learning by proposing Delayed Random Partial Gradient Averaging (DPGA), which reduces transmission size and enables parallel computation, resulting in enhanced system performance as demonstrated on non-IID CIFAR-10/100 datasets.
Federated learning (FL) is a distributed machine learning paradigm that enables multiple clients to train a shared model collaboratively while preserving privacy. However, the scaling of real-world FL systems is often limited by two communication bottlenecks:(a) while the increasing computing power of edge devices enables the deployment of large-scale Deep Neural Networks (DNNs), the limited bandwidth constraints frequent transmissions over large DNNs; and (b) high latency cost greatly degrades the performance of FL. In light of these bottlenecks, we propose a Delayed Random Partial Gradient Averaging (DPGA) to enhance FL. Under DPGA, clients only share partial local model gradients with the server. The size of the shared part in a local model is determined by the update rate, which is coarsely initialized and subsequently refined over the temporal dimension. Moreover, DPGA largely reduces the system run time by enabling computation in parallel with communication. We conduct experiments on non-IID CIFAR-10/100 to demonstrate the efficacy of our method.