Neural gradient boosting in federated learning for hemodynamic instability prediction: towards a distributed and scalable deep learning-based solution.