Modern Challenges in Data Decentralization: Federated Learning, Differential Privacy and Communication Constraints

Overview

Data explosion, characterized by an increase in both volume and complexity, often leads to data decentralization. This decentralization is partly implemented to facilitate storage and enhance computational feasibility.

Decentralized data lead to the possibility of protecting data owners’ privacy but also brings new challenges in how to efficiently communicate between different data sites and how to aggregate knowledge learned from different sites with potential data heterogeneity. In this workshop, we aim to gather researchers from mathematics, statistics, machine learning, computer science, and application areas, to collectively discuss the caveats of the state-of-the-art methods and theory in decentralized learning, as well as to open new possibilities with great minds from different fields.