Federated Learning Bias Mitigation

FedUV: Addressing Bias in Federated Learning with Regularization Techniques

The author introduces FedUV, a method to address bias in federated learning by promoting variance and hyperspherical uniformity in local models to emulate an IID setting.