Learning Reconfigurable Representations for Multimodal Federated Learning with Missing Data

Neural Information Processing Systems 

Multimodal federated learning in real-world settings often encounters incomplete and heterogeneous data across clients.