Dec 7, 2022 · We propose a statistically principled probability-corrected loss to align the posterior probability when models are trained on heterogeneous clients.
Aligning model outputs for class imbalanced non-IID federated learning · Lan Li, De-chuan Zhan, Xin-Chun Li · Published in Machine-mediated learning 7 December ...
Addressing class imbalance in federated learning is crucial for optimal model performance. Learn why it matters and how to tackle it effectively.
Mar 15, 2023 · Abstract—The label distribution skew induced data heterogeniety has been shown to be a significant obstacle that limits the model.
Model Convergence: Models can struggle to converge or might converge to a suboptimal solution when a significant class imbalance exists across clients. Some ...
In this study, we address the model aggregation challenge within FL by focusing on elevating the performance of the global model amidst class imbalance.
May 13, 2024 · We propose a novel method called FedDSE, which can reduce the conflicts among clients by extracting sub-models based on the data distribution of each client.
Jul 23, 2024 · We propose a novel Semi-supervised SFL system, termed SemiSFL, which incorporates clustering regularization to perform SFL with unlabeled and non-IID client ...
The results show that the models trained with non-IID data have consistently lower feature similarity across clients for all layers, compared with those ...
KSAS is designed to sample universally informative data by computing the intensified discrepancies between the clients' and the global model's outputs based on ...