Riemannian Douglas–Rachford Algorithms for Federated Learning and Privacy-Preserving Healthcare Analytics

The aim of this project is to study a data-preserving federated learning approach by designing a Riemannian composite optimization model and developing a Riemannian Doulgas-Rachford algorithm for solving the model. In addition to novel contributions to the filed of optimization and machine learning, our proposed model and algorithm are applied in data-preserving healthcare operations where sharing raw data across various sites is not practical or allowed. This project develops new ways for researchers decision-makers to learn from health data together without sharing any patient records. The work is a joint effort between UBC Okanagan and Eindhoven University of Technology. It will produce theory, modeling, and simulations that both teams can use to support future research, grant applications, and partnerships with healthcare providers. In the long term, this collaboration will help both institutions build stronger expertise in trustworthy artificial intelligence for healthcare and train highly skilled students in this growing area.

Faculty Supervisor:

Amir Ardestani-Jaafari

Student:

Partner:

Eindhoven University of Technology

Discipline:

Mathematics

Sector:

Education

University:

The University of British Columbia - Okanagan

Program: