Medical data sets are difficult to obtain due to the complicated nature of the healthcare system. For example, different hospitals may only be able to access the clinical records of their own patient populations. An intuitive idea is to learn those data without sharing. Federated learning (FL) is a learning paradigm to address the problem of data governance and privacy by learning models collaboratively without exchanging the data.
This project will explore the federated learning in healthcare, e.g., medical image analysis. Several topics will be investigated under FL setting, such as domain generalization, differential privacy, and personalization.
Students are encouraged to publish high-impact research papers.