Federated learning makes it possible for AI algorithms to gain experience from a vast range of data located at different sites. The approach enables several organizations to collaborate on the development of models without directly sharing sensitive data. Preserving privacy while still benefiting from diverse datasets across various locations worldwide.
Today, we are announcing a collaboration between two of the most widely used solutions for federated learning, NVIDIA FLARE and Flower to establish compatibility between the two frameworks. With this collaboration, NVIDIA will support running Flower projects on FLARE, thus enabling FLARE developers to access the rich Flower ecosystem, and give Flower developers new options for production deployment through FLARE. The developer communities of both frameworks will be able to build larger-scale consortiums that leverage more data towards training stronger AI models. Flower is a member of NVIDIA Inception, a program designed to help startups of all stages accelerate growth and innovation.
Flower is known for its ease of use, mobile device support, and a large active open-source community of AI developers and researchers that keep Flower at the forefront of new methods. NVIDIA FLARE offers a comprehensive range of federated learning features including robust communication, concurrent job scheduling, security and confidential computing with strong support for industry-leading NVIDIA hardware. Compatibility between the two complementary frameworks will allow developers to combine the strengths of each platform; under this collaboration, users will be able to build varieties of federated AI systems that were not previously possible.
We introduced this initiative between Flower and NVIDIA at the Flower AI Summit 2024. This event has grown to be the world’s largest conference on federated learning. During the opening AI Industry Day keynote, a live demonstration was provided to the audience of the first working prototype to result from this collaboration. This presentation showed NVIDIA FLARE running an unmodified Flower project that performs federated learning and federated evaluation over several training rounds. This feature is planned to be released to the public with an upcoming NVIDIA FLARE release. We expect that this initial release will be the first of a series of integrations that provide more comprehensive forms of compatibility between Flower and NVIDIA FLARE in the future.