Parameterized Explainer for Graph Neural Network

Authors: Dongsheng Luo, Wei Cheng, Dongkuan Xu, Wenchao Yu, Bo Zong, Haifeng Chen, Xiang Zhang

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on both synthetic and real-life datasets show highly competitive performance with up to 24.7% relative improvement in AUC on explaining graph classification over the leading baseline.
Researcher Affiliation Collaboration Dongsheng Luo1 Wei Cheng2 Dongkuan Xu1 Wenchao Yu2 Bo Zong2 Haifeng Chen2 Xiang Zhang1 1The Pennsylvania State University 2NEC Labs America 1{dul262,dux19,xzz89}@psu.edu 2{weicheng,wyu,bzong,haifeng}@nec-labs.com
Pseudocode Yes Detailed algorithms can be found in the Appendix.
Open Source Code Yes The code and data used in this work are available 2. [Footnote 2: https://github.com/flyingdoog/PGExplainer]
Open Datasets Yes We follow the setting in GNNExplainer and construct four kinds of node classification datasets, BA-Shapes, BA-Community, Tree-Cycles, and Tree-Grids [53]. Furthermore, we also construct a graph classification datasets, BA-2motifs... We also include a real-life dataset, MUTAG, for graph classification, which is also used in previous work [53].
Dataset Splits No The paper states 'We follow the experimental settings in GNNExplainer [53]' but does not explicitly provide the specific percentages or counts for train/validation/test splits for all datasets within the main text.
Hardware Specification No No explicit hardware specifications (e.g., specific GPU or CPU models) used for running the experiments were provided in the paper.
Software Dependencies No No specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions) were provided in the paper.
Experiment Setup No The paper states 'We refer readers to the Appendix for more training details' and mentions tuning temperature τ, but it does not provide specific hyperparameter values (e.g., learning rate, batch size, number of epochs) or detailed optimizer settings within the main text.