FedMD: Heterogenous Federated Learning via Model Distillation
This addresses the need for federated learning in applications like healthcare and AI services where participants have intellectual property concerns and heterogeneous tasks, though it is incremental as it builds on existing distillation and transfer learning techniques.
The paper tackles the problem of federated learning with heterogeneous models, where each participant uses a uniquely designed model, by proposing FedMD, a framework using transfer learning and knowledge distillation. The result shows that with 10 participants, models achieve an average 20% accuracy gain over non-collaborative baselines on MNIST/FEMNIST and CIFAR10/CIFAR100 datasets, nearing performance with pooled data.
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.