Distributionally Robust Federated Learning with Multi-Source Data
Making federated learning work when data is messy and unpredictable
Federated learning trains shared models using private data from many sources, but real-world data often differs wildly between clients in ways that are hard to predict. This paper shows how to build more robust models by letting each client have its own uncertainty buffer rather than forcing one worst-case scenario across everyone—and proves the approach actually works on unseen data.
Federated learning powers systems like predictive keyboards and health research networks where data can't leave users' devices. When these systems fail to handle messy real-world data distributions, they make poor predictions for certain groups of users. This work makes federated learning more reliable when you can't afford to collect perfect statistics upfront, which is common in privacy-sensitive domains where data is limited.