SECAI Core · phase 2 of 15
Federated learning
A machine learning approach where a central model is trained collaboratively across many decentralized devices or servers using their local data, so the raw data never leaves the devices—only model updates (like gradients or parameters) are shared and aggregated to improve the global model while enhancing privacy.
The Explain card
- Plain English
- Federated learning trains one shared model across many devices or servers. Each participant trains on its own local data and sends back only model updates, never the raw data.
- Example
- Thousands of phones each improve a keyboard's next-word prediction locally and send gradient updates to a central server that merges them into the global model.
- Why it matters
- Raw data stays home, which helps privacy, but updates can still leak information about it, and a malicious participant can send poisoned updates to corrupt or backdoor the global model.
- Hook
- The model travels to the data, not the data to the model.
Where it sits in the deck
Phase 2: How Models Are Built: Architectures and Learning Mechanics
With the paradigms named, zoom in on the architectural building blocks that turn data into trained artifacts.
