“In the realm of distributed machine learning, a novel solution called Federated Dual-Decoupling via Model and Logit Calibration (FedDDC) has been introduced to address the challenges of non-IID and long-tailed distributions in federated learning. This approach, characterized by decoupling the global model, client confidence re-weighting, classifier re-balancing, and decoupled knowledge distillation, significantly enhances the accuracy of the global model on non-IID and long-tailed data, outperforming existing state-of-the-art methods.”