Federated Learning with Label Distribution Skew via Logits Calibration

Authors: Jie Zhang, Zhiqi Li, Bo Li, Jianghe Xu, Shuang Wu, Shouhong Ding, Chao Wu

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on federated datasets and real-world datasets demonstrate that Fed LC leads to a more accurate global model and much improved performance.
Researcher Affiliation Collaboration 1Zhejiang University, China 2Youtu Lab, Tencent, China.
Pseudocode Yes The complete pseudo code of Fed LC can be found in Appendix 1. Algorithm 1 Federated Learning via Logits Calibration (for Client)
Open Source Code No The paper does not contain an explicit statement or a link indicating the availability of its source code.
Open Datasets Yes In this study, we conduct a number of experiments on popular image classification benchmark datasets: SVHN (Netzer et al., 2011), CIFAR10 (Krizhevsky et al., 2009), CIFAR100 (Krizhevsky et al., 2009) and Image Net (Deng et al., 2009), as well as federated datasets (Synthetic dataset and FEMNIST) proposed in LEAF (Caldas et al., 2019).
Dataset Splits No The paper discusses training data and testing data but does not explicitly mention using a separate validation split or how such a split was created or accessed.
Hardware Specification Yes all experiments are conducted with 8 Tesla V100 GPUs.
Software Dependencies No The paper states "We implement the typical federated setting (Mc Mahan et al., 2017) in Pytorch," but it does not specify the version number for Pytorch or any other software dependency.
Experiment Setup Yes The size of local mini-batch is 128. For local training, each client updates the weights via SGD optimizer with learning rate η = 0.01 without weight decay. We run each experiment with 5 random seeds and report the average and standard deviation.