We are generating a wealth of data on our personal devices, in smart homes and cities, and within organizations such as hospitals or financial institutions. However, such data is often siloed---residing in the devices or organizations that generated it. Learning across siloed data presents privacy and security risks, and fundamentally changes data analyses from classical scenarios, where we view data as a sample from a single large underlying population. Techniques for federated and collaborative learning aim to enable accurate, trustworthy learning between multiple parties and across heterogeneous data sources. This workshop will explore recent advances and future theoretical directions in federated & collaborative learning, including talks by experts on these topics, as well as focused discussion and brainstorming sessions.
Co-Organizers: John Duchi (Stanford University), Nika Haghtalab (UC Berkeley), Peter Kairouz (Google), Virginia Smith (Carnegie Mellon University), Nati Srebro (TTIC), Kunal Talwar (Apple)