Search papers, labs, and topics across Lattice.
This paper introduces FedReLa, a novel approach for addressing the challenges of class imbalance and data heterogeneity in federated learning by implementing a feature-dependent label re-allocator that adjusts local labels without needing the global class distribution. The method effectively corrects biased global decision boundaries, leading to significant improvements in the accuracy of minority classes and overall model performance on imbalanced datasets. Extensive experiments demonstrate that FedReLa outperforms existing state-of-the-art methods while maintaining low communication overhead, making it a practical solution for decentralized training scenarios.
FedReLa achieves significant accuracy gains for minority classes in federated learning without requiring knowledge of global class distributions.
Federated learning has emerged as the foremost approach for decentralized model training with privacy preservation. The global class imbalance and cross-client data heterogeneity naturally coexist, and the mismatch between local and global imbalances exacerbates the performance degradation of the aggregated model. The agnosticism of global class distribution poses significant challenges for data-level methods, especially under extreme conditions with severe class absence across clients. In this paper, we propose FedReLa, a novel data-level approach that tackles the coexistence of data heterogeneity and class imbalance in federated learning. By re-labeling samples with a feature-dependent label re-allocator, FedReLa corrects biased global decision boundaries without requiring knowledge of the global class distribution. This modular, model-agnostic approach can be integrated with algorithmic methods to deliver consistent improvements without additional communication overhead. Through extensive experiments, our method significantly improves the accuracy of minority classes and the overall accuracy on stepwise-imbalanced and long-tailed datasets, outperforming the previous state of the art.
