Power up! Robust Graph Convolutional Network via Graph Powering
Authors: Ming Jin, Heng Chang, Wenwu Zhu, Somayeh Sojoudi8004-8012
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The proposed approaches are demonstrated in extensive experiments to simultaneously improve performance in both benign and adversarial situations. ... We followed the setup of (Yang, Cohen, and Salakhutdinov 2016; Kipf and Welling 2017) for citation networks Citeseer, Cora and Pubmed (please refer to the Appendix for more details). ... Model Citeseer Cora Pubmed ... Table 1: Results for semi-supervised node classification. |
| Researcher Affiliation | Academia | 1 Tsinghua-Berkeley Shenzhen Institute, Tsinghua University 2 Department of Electrical and Computer Engineering, Virginia Tech 3 Department of Computer Science and Technology, Tsinghua University 4 Department of Electrical Engineering and Computer Sciences, University of California at Berkeley |
| Pseudocode | No | The paper describes its methods and algorithms using mathematical formulations and textual descriptions, but it does not include any explicitly labeled pseudocode blocks or algorithm listings. |
| Open Source Code | No | The paper mentions that "All baseline methods on based on their public codes" but does not provide any statement or link for the code implementation of the proposed method. |
| Open Datasets | Yes | We followed the setup of (Yang, Cohen, and Salakhutdinov 2016; Kipf and Welling 2017) for citation networks Citeseer, Cora and Pubmed (please refer to the Appendix for more details). |
| Dataset Splits | Yes | The label rates are 0.1%, 0.5% and 1%, the validation rate is 1%, and the rest of the nodes are testing points. |
| Hardware Specification | No | The paper does not provide any specific details regarding the hardware (e.g., CPU, GPU models, memory, or specific computing platforms) used for conducting the experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., Python version, library versions like PyTorch 1.x, TensorFlow 2.x) needed to replicate the experiments. It mentions using "public codes" for baselines but not for its own implementation details. |
| Experiment Setup | Yes | Graph powering order can influence spatial and spectral behaviors. Our theory suggests powering to the order of log(n); in practice, orders of 2 to 4 suffice (Figure 6). Here, we chose the power order to be 4 for r-GCN on Citeseer and Cora, and 3 for Pubmed, and reduced the order by 1 for VPN. |