Geometric Knowledge Distillation: Topology Compression for Graph Neural Networks

Authors: Chenxiao Yang, Qitian Wu, Junchi Yan

NeurIPS 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conduct experiments to validate the efficacy of our method on graph-structured data in terms of various types of privileged geometric knowledge, combinations of teacher-student GNN architectures and potential application scenarios. We use three benchmark datasets Cora [35], Citeseer [44], Pubmed [39], and a larger dataset OGBN-Arxiv [23] for node classification tasks.
Researcher Affiliation Academia Chenxiao Yang, Qitian Wu, Junchi Yan Department of Computer Science and Engineering Mo E Key Lab of Artificial Intelligence Shanghai Jiao Tong University {chr26195,echo740,yanjunchi}@sjtu.edu.cn
Pseudocode No The paper describes algorithmic steps in prose but does not include a formally labeled pseudocode or algorithm block.
Open Source Code Yes The codes are available at https://github.com/chr26195/GKD.
Open Datasets Yes We use three benchmark datasets Cora [35], Citeseer [44], Pubmed [39], and a larger dataset OGBN-Arxiv [23] for node classification tasks.
Dataset Splits No The paper mentions 'best validation accuracy' (Figure 4 caption) implying a validation set was used, and confirms 'data splits' were specified in the checklist (3b), but specific percentages or methods for creating these splits are not explicitly detailed in the main text.
Hardware Specification No The paper does not provide specific hardware details such as GPU/CPU models, types of processors, or cloud instance specifications used for running experiments.
Software Dependencies No The paper does not provide specific version numbers for software dependencies or libraries used in the implementation.
Experiment Setup Yes The backbone f is set as 3-layer GCN [28] for both student and teacher models, unless otherwise stated.