A Universal Federation of a Mixture of Highly Heterogeneous Learning Tasks | MICCAI MLMI 2024

#DeepLearning #FederatedLearning #MICCAI2024

👉 Paper link: https://link.springer.com/chapter/10.1007/978-3-031-73290-4_4
👉 arXiv link: TO-UPDATE
👉 Code link: https://github.com/basiralab/UniFed

UniFed is a novel federated learning framework that learns from a mixture of highly heterogeneous tasks by introducing a loss-guided dynamic and sequential model exchange between the server and clients with application to medical image classification. We further remedy the variance in convergence time across clients during our sequential exchange process while being agnostic to task complexity.

Скачать видео