Federated learningΒΆ

Federated learning allows for multiple clients each with their own data to collaborate on training together without having to share their actual data.

NVFlare (NVIDIA Federated Learning Application Runtime Environment) is now a more generalized package that can be used in contexts outside of Clara as well. It contains the core code to make federated learning run, the provisioning tool, and now new homomorphic encryption tools to allow you to compute data while the data is still encrypted. See the NVFlare API reference below for details.

Despite the transition to NVFlare, the functionality of federated learning in Clara 4.0 is mostly similar to how it was back in Clara 3.1 with the same provisioning, starting, and operating process, but with some configuration changes.