Loading timeline…
20162010s
Federated Learning
Training a centralized AI model using decentralized data stored on millions of user devices.
Why It Was Important
Introduced by Google Research, this technique enables edge devices (like smartphones) to cooperatively learn a shared prediction model while keeping all training data perfectly localized. It drastically improves privacy by never transmitting user data—only the learned gradients—to the central server.
Who Invented It
Brendan McMahan et al.
Privacy-focused cryptography and ML researchers.
Applications
- Keyboard Prediction (Gboard)
- Healthcare AI (HIPAA)
- Privacy-preserving ML
Key Papers
- Communication-Efficient Learning of Deep Networks from Decentralized Data
H. Brendan McMahan et al. · AISTATS 2017 · 2016
Videos
Training AI Models with Federated Learning
IBM Technology
Stanford Seminar - Federated Learning in Medicine: Breaking Down Silos to Advance Medical Research
Stanford Online