FedCD: Improving Performance in non-IID Federated Learning

TL;DR

FedCD is a novel federated learning method that dynamically groups devices with similar data by cloning and deleting models, leading to improved accuracy and convergence on non-IID data with minimal overhead.

Contribution

The paper introduces FedCD, a new approach that enhances federated learning on non-IID data by dynamically clustering devices based on data similarity.

Findings

FedCD outperforms FedAvg in accuracy on CIFAR-10 non-IID data.

FedCD converges faster than baseline methods.

FedCD incurs minimal additional computational and communication costs.

Abstract

Federated learning has been widely applied to enable decentralized devices, which each have their own local data, to learn a shared model. However, learning from real-world data can be challenging, as it is rarely identically and independently distributed (IID) across edge devices (a key assumption for current high-performing and low-bandwidth algorithms). We present a novel approach, FedCD, which clones and deletes models to dynamically group devices with similar data. Experiments on the CIFAR-10 dataset show that FedCD achieves higher accuracy and faster convergence compared to a FedAvg baseline on non-IID data while incurring minimal computation, communication, and storage overheads.