Structure-free Graph Condensation: From Large-scale Graphs to Condensed Graph-free Data

Authors: Xin Zheng, Miao Zhang, Chunyang Chen, Quoc Viet Hung Nguyen, Xingquan Zhu, Shirui Pan

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 3 Experiments; Table 1: Node classification performance (ACC% ± std) comparison between condensation methods and other graph size reduction methods with different condensation ratios.
Researcher Affiliation Academia 1Monash University, Australia, 2Harbin Institute of Technology (Shenzhen), China 3Griffith University, Australia, 4Florida Atlantic University, USA
Pseudocode Yes Algorithm 1 Structure-Free Graph Condensation (SFGC)
Open Source Code Yes Corresponding author Code is available at https://github.com/Amanda-Zheng/SFGC
Open Datasets Yes Following [27], we evaluate the node classification performance of the proposed SFGC method on Cora, Citeseer [61], and Ogbn-arxiv [17] under the transductive setting, on Flickr [66] and Reddit [16] under the inductive setting. ... The dataset statistic details are shown in Appendix C.
Dataset Splits Yes For all datasets under two settings, we use the public splits and setups for fair comparisons. ... The dataset statistic details are shown in Appendix C. (Table A1 provides specific numbers for train/val/test splits).
Hardware Specification Yes Table A3: Running time comparison (seconds) of the proposed SFGC and GCOND [27] for 50 epochs with a single GeForce RTX 3080 GPU.
Software Dependencies No The paper mentions using the “GCN model” and “two-layer GNN” but does not specify software dependencies with version numbers (e.g., PyTorch 1.x, Python 3.x).
Experiment Setup Yes Additional hyper-parameter setting details are listed in Appendix E. (Appendix E, Table A5 provides specific hyperparameter values).