Speaker: Wotao Yin, University of California, Los Angeles

Title: A Method for Vertical Federated Learning

Abstract: Federated learning (FL) is a form of distributed learning in which training data are scattered and must be secured by local agents. FL is to enable model training with multi-agent data where privacy and data security prevent us from using the existing machine learning algorithms. Besides security, agents in the FL setting typically have heterogeneous capacities of computing powers and network connections. This talk considers the scenario where different agents have different data describing the same set of subjects, also known as Vertical FL. A method that integrates asynchronous block-coordinate descent, differential privacy, and local embedding is presented. The method has provable convergence performance along with data privacy guarantees. It has the potential to facilitate the collaborations of banks, hospitals, and cross-domain business to improve their services for their joint customers. This talk is based on a joint work with Tianyi Chen and Xiao Jin at RPI and Yuejiao Sun at UCLA.

Department of Mathematical Sciences
George Mason University
4400 University Drive, MS 3F2
Fairfax, VA 22030-4444
http://math.gmu.edu/
Tel. 703-993-1460, Fax. 703-993-1491