Federated learning on non-IID data: A survey

Federated learning is an emerging distributed machine framework for privacy preservation. However, models trained in federated usually have worse performance than those the standard centralized mode, especially when training data are not independent and identically (Non-IID) on local devices. In this survey, we provide a detailed analysis of influence Non-IID both parametric non-parametric horizontal vertical learning. addition, current research work handling challenges reviewed, advantages disadvantages these approaches discussed. Finally, suggest several future directions before concluding paper.