Graph Neural Networks Inspired by Classical Iterative Algorithms
Authors: Yongyi Yang, Tang Liu, Yangkun Wang, Jinjing Zhou, Quan Gan, Zhewei Wei, Zheng Zhang, Zengfeng Huang, David Wipf
ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We test the performance of TWIRLS under different application scenarios, in each case comparing against SOTA methods specifically designed for the particular task at hand. |
| Researcher Affiliation | Collaboration | 1Fudan University 2Shanghai Jiao Tong University 3Amazon 4Renmin University of China. |
| Pseudocode | No | The paper describes iterative algorithms using equations and textual descriptions but does not provide structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our code is available at this link. |
| Open Datasets | Yes | three commonly used citation datasets, namely Cora, Pubmed and Citeseer (Sen et al., 2008), plus ogbn-arxiv (Hu et al., 2020), a relatively large dataset. and We select four graph datasets with a low homophily ratio, namely, Actor, Cornell, Texas, and Wisconsin, adopting the data split, processed node features, and labels provided by (Pei et al., 2019). |
| Dataset Splits | Yes | For the citation datasets, we use the semi-supervised setting from (Yang et al., 2016), while for ogbn-arxiv, we use the standard split from the Open Graph Benchmark (OGB) leaderboard (see https://ogb.stanford.edu). and adopting the data split, processed node features, and labels provided by (Pei et al., 2019). |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions that 'All models and experiments were implemented using the Deep Graph Library (DGL) (Wang et al., 2019)', but does not provide specific version numbers for DGL or other software dependencies. |
| Experiment Setup | No | The paper states: 'For full details regarding the propagation steps, attention formulation, and other hyperparameters, please see the supplementary; similarly for experimental details, additional ablation studies and empirical results, and a discussion of computational complexity.' This indicates that specific experimental setup details are not provided in the main text. |