UCLA Mathematics Professor Wotao Yin‘s paper titled, “Hybrid Federated Learning: Algorithms and Implementation,” won the Best Student Paper Award at the 34th Conference on Neural Information Processing Systems (NeurIPS-20) Workshop on Scalability, Privacy, and Security in Federated Learning. The paper, written in collaborations with researchers from the University of Minnesota and Rensselaer Polytechnic Institute, was published on February 18, 2021 on arXiv.org.
Yin explains, “The research is about federated learning, a new form of machine learning where the data from different participants are subject to distributed processing and privacy requirements. They jointly do machine learning without directly exchanging their data. In particular, the proposed method works when each participant has the records of only a portion of all the subjects and their records are possibly highly incomplete. As a potential application, a union of loan and insurance providers can work together to build a model to compute a customer’s credit without sharing or exchanging their customers’ credit histories. (On the contrary, today, they report everyone’s full credit histories to credit bureaus, and a lender must access your full history during your loan application. This is both unsafe and expensive.)”
