Robust Training of Federated Models with Extremely Label Deficiency
Authors: Yonggang Zhang, Zhiqin Yang, Xinmei Tian, Nannan Wang, Tongliang Liu, Bo Han
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our comprehensive experiments on four benchmark datasets provide substantial evidence that Twin-sight can significantly outperform state-of-the-art methods across various experimental settings, demonstrating the efficacy of the proposed Twin-sight. |
| Researcher Affiliation | Academia | Yonggang Zhang1 Zhiqin Yang1 Xinmei Tian2 Nannan Wang3 Tongliang Liu4 Bo Han1 1TMLR Group, Hong Kong Baptist University 2University of Science and Technology of China 3Xidian University 4Sydney AI Centre, The University of Sydney |
| Pseudocode | Yes | Algorithm 1 pseudo-code of Twin-sight |
| Open Source Code | Yes | The code is publicly available at: github.com/tmlr-group/Twin-sight. |
| Open Datasets | Yes | In our experiments, we use four popular datasets that have been extensively utilized in FSSL research (Liang et al., 2022; Wei & Huang, 2023) including CIFAR-10 (Krizhevsky et al., 2009), SVHN (Netzer et al., 2011), Fashion-MNIST (FMNIST) (Xiao et al., 2017), and CIFAR-100 (Krizhevsky et al., 2009). |
| Dataset Splits | Yes | The training sets of these four datasets are 50, 000, 73, 257, 60, 000, and 50, 000 respectively. They are partitioned into K clients in federated learning, and we resize all images to 32 32 size. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU models, CPU types) used for running the experiments. |
| Software Dependencies | No | The paper mentions using Resnet-18 as a backbone feature extractor and SGD optimizer, but does not list specific software dependencies with version numbers (e.g., PyTorch 1.x, TensorFlow 2.x). |
| Experiment Setup | Yes | We use the SGD optimizer with a learning rate of 0.01, weight decay of 0.0001, and momentum of 0.9 in all of our experiments. The batch size is set to 64 for all datasets. |