FedCL: Federated Multi-Phase Curriculum Learning to Synchronously Correlate User Heterogeneity. (arXiv:2211.07248v1 [cs.LG] CROSS LISTED)

Federated Learning (FL) is a new decentralized learning used for training
machine learning algorithms where a global model iteratively gathers the
parameters of local models but does not access their local data. A key
challenge in FL is to handle the heterogeneity of local data distribution,
resulting in a drifted global model, which is hard to converge. To cope with
this challenge, current methods adopt different strategies like knowledge
distillation, weighted model aggregation, and multi-task learning, as
regulation. We refer …

arxiv curriculum curriculum learning