Dr. Eng Lim Goh, Joachim Schultze, & Krishna Prasad Shastry | HPE Discover 2020
summary
Describes a working swarm learning system that averages learned neural network weights across hospitals while keeping raw patient data local, using averaging learned neural network weights and blockchain for decentralized coordination to remove central servers. Implements containerized microservices, REST APIs and a ten‑layer neural net example converted with fewer than ten lines of code. Handles non‑IID data, node dropouts and differing lab technologies; applied to leukemia and COVID blood transcriptomes across multi‑site networks (10 sites in Germany) and shows improved, less biased accuracy.
AI-written summary · about
chapters (18)
- 0:00Introduction and guests
- 1:48Global collaboration in pandemic response
- 2:31Swarm learning fundamentals
- 5:01Precision medicine applications
- 6:49Technical implementation details
- 9:18COVID-19 research use case
- 11:05Data privacy and bias removal
- 13:52EU privacy compliance importance
- 15:39Continuous learning in medical AI
- 17:02Simplifying swarm learning adoption
- 18:26Real-world implementation progress
- 20:12Interdisciplinary team requirements
- 21:41Technology bias handling
- 23:49Clinical trial optimization
- 25:37Edge AI and swarm learning
- 27:23Future possibilities and vision
- 30:56Democratization of AI
- 32:25Tech for good and closing
