Federated Learning: Unlocking the Values of Discrete Data for Enterprise
summary
Overview of production federated learning that cites two open source stacks—FATE and OpenFL—and tooling to run them across clouds including CubeFate and FedLCM. Claims: community of 4,000+ engineers, 40+ releases, 30+ built‑in algorithms; FATE LLM supports distributed fine‑tuning of LLMs using adapter fine-tuning that cuts tuned parameters to ~1–10% and training time ~80% savings. Notes tradeoffs: homomorphic encryption imposes ~10×–50× slowdown. Shows examples with ~30–50 hospital participants and describes aggregation, encryption, and lifecycle management.
AI-written summary · about
chapters (11)
- 0:00Start
- 2:37What is Federated Learning?
- 9:16Categories of Federated Learning
- 12:19Federated Learning Frameworks in LF AI & Data
- 15:24FATE Framework
- 22:53FATE-LLM high-level architecture
- 25:35OpenFL high-level architecture (Cont.)
- 26:35OpenFL Case Study
- 27:12Data ecosystem
- 28:29Interoperability between FL frameworks
- 32:17Federated Learning from Operator's view
related talks
21mWhat, Why and How of Federated Machine Learning – Implementation Using...
31mFrom Theory to Practice: Advances in Real-World Federated Learning
25mData collaboration with Federated Learning
41mFederated Learning: Machine Learning on Decentralized Data
27mCreate Value and Insight Through Privacy-preserving Data Collaboration [...]
31mFedLess: Secure and Scalable Serverless Federated Learning
31mFederated Learning: So Close and Yet So Out of Reach
29mBalancing Thin Line Between Data Intelligence & Privacy