On Privacy and Personalization in Federated Learning: A Retrospective on the US/UK PETs Challenge
ML @ CMU
MAY 12, 2023
As we discuss below, this alternate privacy granularity affects how we consider modeling federated data to improve privacy/utility trade-offs. It is also very easy to implement: each silo can just run DP-SGD for local gradient steps with calibrated per-step noise. non-identically distributed data) exists between data silos.
Let's personalize your content