Understanding Expressivity of GNN in Rule Learning
Authors: Haiquan Qiu, Yongqi Zhang, Yong Li, quanming yao
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results are consistent with our theoretical findings and verify the effectiveness of our proposed method. The code is publicly available at https: //github.com/LARS-research/Rule-learning-expressivity. In this section, we validate our theoretical findings from Section 3 and showcase the efficacy of our proposed EL-GNN (Section 4) on synthetic and real datasets through experiments. |
| Researcher Affiliation | Academia | Haiquan Qiu1, Yongqi Zhang2, Yong Li1, Quanming Yao1 1Department of Electronic Engineering, Tsinghua University 2The Hong Kong University of Science and Technology (Guangzhou) |
| Pseudocode | Yes | Algorithm 1 Entity Labeling |
| Open Source Code | Yes | The code is publicly available at https: //github.com/LARS-research/Rule-learning-expressivity. |
| Open Datasets | Yes | We follow the standard setup as Zhu et al. (2021) to test EL-GNN s effectiveness on four real datasets: Family (Kok & Domingos, 2007), Kinship (Hinton et al., 1986), UMLS (Kok & Domingos, 2007), WN18RR (Dettmers et al., 2017), and FB15k-237 (Toutanova & Chen, 2015). |
| Dataset Splits | Yes | Triplets with the target relation are separated into training, validation, and testing sets. Table 1: Accuracy on synthetic data. Table 6: Statistics of the synthetic datasets. |
| Hardware Specification | Yes | All experiments were implemented in Python using Py Torch and executed on A100 GPUs with 80GB memory. |
| Software Dependencies | No | The paper states that experiments were "implemented in Python using Py Torch" but does not provide specific version numbers for either Python or Py Torch, or any other relevant software libraries. |
| Experiment Setup | Yes | Hyperparameters for all methods are automatically tuned with Ray (Liaw et al., 2018) based on the validation accuracy. For real datasets, we uses d = 5, 30, 100, 100, 300 for Family, Kinship, UMLS, WN18RR, and FB15k-237, respectively. |