AllMatch: Exploiting All Unlabeled Data for Semi-Supervised Learning
Authors: Zhiyu Wu, Jinshi Cui
IJCAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We extensively evaluate our approach on multiple benchmarks, encompassing both balanced and imbalanced settings. The results demonstrate that All Match consistently outperforms existing state-of-the-art methods. |
| Researcher Affiliation | Academia | National Key Laboratory of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University wuzhiyu@pku.edu.cn, cjs@cis.pku.edu.cn |
| Pseudocode | No | The paper describes its methods in text and figures, but it does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | No | The paper states 'we employ the unified codebase Torch SSL [Zhang et al., 2021] to evaluate all methods' but does not provide a link or explicit statement about the availability of their own specific source code for 'All Match'. |
| Open Datasets | Yes | For balanced image classification, we conduct experiments on CIFAR-10/100 [Krizhevsky et al., 2009], SVHN [Netzer et al., 2011], STL-10 [Coates et al., 2011], and Image Net [Deng et al., 2009] with various numbers of labeled data, where the class distribution of the labeled data is balanced. |
| Dataset Splits | No | The paper mentions 'labeled and unlabeled datasets' and evaluation on 'test' data, but it does not explicitly describe the use or definition of a separate 'validation' dataset split for hyperparameter tuning or early stopping. |
| Hardware Specification | No | The paper mentions backbone architectures like WRN-28-2 and ResNet-50 but does not provide any specific details about the hardware (e.g., GPU models, CPU types, memory) used to run the experiments. |
| Software Dependencies | No | The paper states 'we employ the unified codebase Torch SSL [Zhang et al., 2021] to evaluate all methods', but it does not specify version numbers for any software dependencies like Python, PyTorch, CUDA, or other libraries. |
| Experiment Setup | Yes | All Match is trained using the SGD optimizer with an initial learning rate of 0.03 and a momentum decay of 0.9. The learning rate is adjusted by a cosine decay scheduler over a total of 220 iterations. We set m to 0.999 and generate the EMA model with a momentum decay of 0.999 for inference. The upper bound K is set to 20 for Image Net and 10 for the other datasets. For SVHN, CIFAR-10 with 10 labels, and STL-10 with 40 labels, we constrain the threshold within the range of [0.9, 1.0] to prevent overfitting noisy pseudo-labels in the early training stages. The batch sizes BL and BU are set to 128 and 128 for Image Net and 64 and 448 for the remaining datasets. For imbalanced SSL, the batch sizes BL and BU are set to 64 and 128, respectively. The learning rate is initially set to 2e-3 and adjusted by a cosine decay scheduler during training. |