FedMD: Heterogenous Federated Learning via Model Distillation

TL;DR

FedMD introduces a federated learning framework that allows heterogeneous models designed independently by participants to collaboratively improve through transfer learning and knowledge distillation, achieving near-centralized performance.

Contribution

This work presents a novel federated learning approach accommodating heterogenous models using transfer learning and knowledge distillation, addressing privacy and model design diversity.

Findings

Achieved 20% average accuracy gain with 10 participants.

Models' performance was only a few percent below pooled data baseline.

Framework demonstrated fast convergence on multiple datasets.

Abstract

Federated learning enables the creation of a powerful centralized model without compromising data privacy of multiple participants. While successful, it does not incorporate the case where each participant independently designs its own model. Due to intellectual property concerns and heterogeneous nature of tasks and data, this is a widespread requirement in applications of federated learning to areas such as health care and AI as a service. In this work, we use transfer learning and knowledge distillation to develop a universal framework that enables federated learning when each agent owns not only their private data, but also uniquely designed models. We test our framework on the MNIST/FEMNIST dataset and the CIFAR10/CIFAR100 dataset and observe fast improvement across all participating models. With 10 distinct participants, the final test accuracy of each model on average receives a 20% gain on top of what's possible without collaboration and is only a few percent lower than the performance each model would have obtained if all private datasets were pooled and made directly available for all participants.