Prototype Similarity Distillation for Communication-Efficient Federated Unsupervised Representation Learning | IEEE Journals & Magazine | IEEE Xplore