Talk by Micah Sheller (Flower/ML Commons)
summary
Cross-silo federated learning runs real experiments across 71 hospitals and yields concrete outcomes: a simple ReSCUNet tumor segmentation model beat state-of-the-art because of diverse data, but setup consumes huge governance and grad‑student effort. Industry use cases like pharmaceutical candidate screening promise billion‑dollar value yet stall on legal, InfoSec and trust barriers. Work centers on building governance as code—algorithms that predict what a workload will learn and automate policy enforcement so iterations become feasible.
AI-written summary · about
chapters (7)
- 0:00Cross-silo federated learning intro
- 3:04Cross-silo vs cross-device FL
- 7:42Medical FL challenges
- 12:37Trust and governance in FL
- 18:32Model value and policy needs
- 24:05Automating governance for FL
- 26:42Q&A
related talks
32mInnovation in the Age of Regulation: Federated Learning with Flower
18mEnabling Medical AI through Federated Learning
03h 51mMachine Learning Summit
45mManaged federated learning on AWS: A case study for healthcare
26mMachine Learning without the Data #federatedlearning
13mScalable and Portable Federated Learning Simulation Engine
22mCollaborative Self-Learning for Security, Fun and Profit
40mDemystifying Decentralized Fair Models for Federated Machine Learning