Open Internet by MindsNet

Addressing Non-IID Data Challenges in Federated Learning

Federated learning faces performance drops and unstable convergence with non-IID data. Client-level variability significantly impacts global model accuracy. Improving model performance in real-world federated settings with naturally non-IID data is challenging.

Computing & Technology, Computer Science, Machine Learning