Self-Training Based Few-Shot Node Classification by Knowledge Distillation
Authors: Zongqian Wu, Yujie Mo, Peng Zhou, Shangbo Yuan, Xiaofeng Zhu
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results showed that our method achieves supreme performance, compared with state-of-the-art methods. Our code and a comprehensive theoretical version are available at https://github.com/zongqianwu/KD-FSNC. Experiments Experimental Settings Datasets and Comparison Methods We assess the effectiveness of our proposed method on six benchmark datasets, including three citation datasets (i.e., Cora, Cite Seer and Cora Full) (Kipf and Welling 2016), two business datasets (i.e., Computers and Photo) (Shchur et al. 2018), and one coauthorship dataset i.e., Coauthor (CS) (Shchur et al. 2018). |
| Researcher Affiliation | Academia | 1School of Computer Science and Engineering, University of Electronic Science and Technology of China, Chengdu, China 2Shenzhen Institute for Advanced Study, University of Electronic Science and Technology of China, Shenzhen, China 3College of Computer Science and Electronic Engineering, Hunan University, Changsha, China 4School of Engineering and Design, Technical University of Munich, Munich, Germany |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | Yes | Our code and a comprehensive theoretical version are available at https://github.com/zongqianwu/KD-FSNC. |
| Open Datasets | Yes | We assess the effectiveness of our proposed method on six benchmark datasets, including three citation datasets (i.e., Cora, Cite Seer and Cora Full) (Kipf and Welling 2016), two business datasets (i.e., Computers and Photo) (Shchur et al. 2018), and one coauthorship dataset i.e., Coauthor (CS) (Shchur et al. 2018). |
| Dataset Splits | No | The paper describes few-shot learning splits into 'base set', 'novel set', 'support set', and 'query set', but does not explicitly mention or provide details for a distinct 'validation set' split or percentages for a traditional train/validation/test split for the entire dataset. |
| Hardware Specification | Yes | All experiments are conducted on a server with NVIDIA Tesla V100S (32GB memory each). |
| Software Dependencies | No | The paper mentions using the Adam optimization algorithm but does not specify any software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | In our proposed method, we optimize all parameters by Adam optimization algorithm (Kingma and Ba 2014) and set the learning rate in the range of {0.01, 0.02, ..., 0.05}. We set the self-training cycle times t in the range of {1, 2, ..., 10}, and set the number m of pseudo-labels in the range of {10, 20, ..., 100}. We vary the values of µ1 and µ2 in the range of {0.1, 0.2, ..., 1.0}. |