CA-MMDS

New framework enables flexible federated learning with evolving clients

Researchers have developed a new framework called CA-MMDS to address challenges in federated learning, particularly for evolving client sets and changing label spaces. This continual multiple-model distillation approach…