English Español Français Deutsch
← Back to the list
Foundational figures Foundational figures Methodologically adjacent

Qiang Yang

Whether useful machine learning can happen when the data you'd train or evaluate on can't be moved to a single place — for legal, privacy, or commercial reasons.

Yang co-authored the foundational textbook on federated learning, and his ongoing research line — first at HKUST and now also through WeBank — is about the same problem at industrial scale: build ML systems where multiple organizations contribute data without the data ever leaving their premises. The transfer-learning textbook before that set out the related question of when knowledge from one task or domain can usefully move to another. For ai100, which currently audits engines on public-query behavior but may one day need to evaluate model behavior on data brand owners can't publish, federated-evaluation methodology is the literature that already worked out the privacy-versus-utility trade-offs.

Worth following when
you need to evaluate or train ML systems on data that has to stay distributed, and want the methodology that the federated-learning community spent a decade developing.
Topics
federated learning foundations; transfer learning across tasks and domains; the privacy-versus-utility trade-offs in distributed ML evaluation.
Key works
"Federated Machine Learning: Concept and Applications" foundational paper (2019); transfer learning textbook (multiple editions); long body of work at HKUST and WeBank on federated AI.