Navigating the privacy-accuracy tradeoff: federated survival analysis with binning and differential privacy
Federated Learning (FL) offers a decentralized approach to model training, allowing for data-driven insights while safeguarding patient privacy across institutions. In the Personal Health Train (PHT) paradigm, it is the local model gradients from each institution, aggregated over a sample size of its own patients that...