Simple and Efficient Heterogeneous Graph Neural Network

Authors: Xiaocheng Yang, Mingyu Yan, Shirui Pan, Xiaochun Ye, Dongrui Fan

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on five real-world heterogeneous graphs demonstrate the superiority of Se HGNN over the state-of-the-arts on both accuracy and training speed.
Researcher Affiliation Academia 1 State Key Lab of Processors, Institute for Computing Technology, Chinese Academy of Sciences, China 2 School of Information and Communication Technology, Griffith University, Australia 3 School of Computer Science and Technology, University of Chinese Academy of Sciences, China {yangxiaocheng, yanmingyu}@ict.ac.cn, s.pan@griffith.edu.au, {yexiaochun, fandr}@ict.ac.cn
Pseudocode Yes Algorithm 1: The overall training process of Se HGNN
Open Source Code Yes Codes are available at https://github.com/ICT-GIMLab/Se HGNN.
Open Datasets Yes Experiments are conducted on four widely-used datasets from HGB benchmark (Lv et al. 2021), as well as a large-scale dataset ogbn-mag from OGB challenge (Hu et al. 2020a).
Dataset Splits Yes All results presented in this section are the average of 20 runs with different data partitions to mitigate the influence of random noise. [...] The ogbn-mag dataset presents two extra challenges: (1) some types of nodes lack raw features, (2) target-type nodes are split according to years, causing training nodes and test nodes to have different data distributions.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory amounts) used for running experiments were mentioned in the paper.
Software Dependencies No No specific software dependencies with version numbers (e.g., Python 3.8, PyTorch 1.9, CUDA 11.1) were mentioned in the paper.
Experiment Setup No The details about all experiment settings and the network configurations are recorded in Appendix1.