NeurIPS 2019 Expo Talk
Dec. 8, 2019
Federated Learning in Healthcare
Philip Dow (Doc.ai)
Machine learning promises to revolutionize many industries, but its application is restricted to areas where there is enough data to train useful models. Often, the barriers to data acquisition are not technological but issues such as privacy, trust, regulatory compliance, and intellectual property. This is especially the case in healthcare, where patients and consumers expect privacy with respect to personal information and where organizations want to protect the value of their data and are also required to follow regulatory laws such as HIPAA in the United States and the GDRP in the Eurozone. Federated learning, which provides the ability to share a model without sharing the data used to train it, has the potential to address these concerns. This talk will explore the application of Federated Learning to problems in healthcare. We’ll examine two applications specifically: Federated Mobile Learning, which takes place in the consumer space where data is located on a user’s personal device, and Federated Cloud Learning, which focuses on business applications in which internal company data cannot be shared with other entities or even within an organization itself. We will discuss the challenges faced by Federated Learning, such as the privacy guarantees a Federated Learning approach can make, and we will look at two of the most common techniques: differential privacy and homomorphic encryption. We will also address some of the engineering and theoretical challenges of Federated Learning. Finally, we will conclude that Federated Learning is a viable approach to machine learning in the healthcare space that can address patient, business, and regulatory concerns with the application of privacy-preserving techniques.