Half-Hop: A graph upsampling approach for slowing down message passing
Authors: Mehdi Azabou, Venkataramana Ganesh, Shantanu Thakoor, Chi-Heng Lin, Lakshmi Sathidevi, Ran Liu, Michal Valko, Petar Veličković, Eva L Dyer
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We report results on several supervised and self-supervised benchmarks, and show improvements across the board, notably in heterophilic conditions where adjacent nodes are more likely to have different labels. |
| Researcher Affiliation | Collaboration | 1Georgia Tech 2Deep Mind. Correspondence to: Mehdi Azabou and Eva Dyer <{mazabou, evadyer}@gatech.edu>. |
| Pseudocode | No | The paper does not contain explicit pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code is provided at: https://github.com/nerdslab/halfhop. |
| Open Datasets | Yes | We use five real-world datasets, Amazon Computers and Amazon Photos (Mc Auley et al., 2015), Coauthor CS and Coauthor Physics (Sinha et al., 2015) and Wiki CS (Mernyei & Cangea, 2020). |
| Dataset Splits | Yes | We follow a 60:20:20% train/val/test split for the Amazon and Coauthor datasets, and 20 pre-split masks provided in the Wiki CS dataset. |
| Hardware Specification | Yes | OOM indicates out-of-memory on a 48GB Nvidia A40 GPU. |
| Software Dependencies | No | The paper mentions "Scikit-learn library (Pedregosa et al., 2011)" but does not provide specific version numbers for software dependencies, which is required for reproducibility. |
| Experiment Setup | Yes | The hyperparameter tuning is performed using the development set only, and the accuracy of the best model on the development set is reported on the test set. For heterophilic datasets, we use the splits provided by (Pei et al., 2020), and also follow the same hyperparameter search protocol. For self-supervised benchmarks, we use the standard hyperparameters provided for each model and dataset (Zhu et al., 2020b; Thakoor et al., 2022). We provide more details in Appendix C. Table 5. Best hyperparameters found using the validation set in our experiments on heterophilic datasets. |