Hierarchical Transformer for Scalable Graph Learning

Authors: Wenhao Zhu, Tianyu Wen, Guojie Song, Xiaojun Ma, Liang Wang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical evaluations demonstrate that HSGT achieves stateof-the-art performance on large-scale benchmarks with graphs containing millions of nodes with high efficiency.
Researcher Affiliation Collaboration 1National Key Laboratory of General Artificial Intelligence, School of Intelligence Science and Technology, Peking University 2Yuanpei College, Peking University 3Microsoft 4Alibaba Group
Pseudocode Yes Algorithm 1 Overview of HSGT
Open Source Code No The paper does not provide concrete access to source code for the methodology described, nor does it explicitly state that the code is publicly available.
Open Datasets Yes We conduct experiments on nine benchmark datasets including four small-scale datasets (Cora, Cite Seer, Pub Med [Sen et al., 2008; Yang et al., 2016], Amazon-Photo [Shchur et al., 2018]) and six large-scale datasets (ogbnarxiv, ogbn-proteins, ogbn-products [Hu et al., 2020], Reddit [Hamilton et al., 2017], Flickr, Yelp [Zeng et al., 2019]).
Dataset Splits Yes We use the predefined dataset split if possible, or we set a random 1:1:8 train/valid/test split.
Hardware Specification No The paper mentions 'GPU memory usage' but does not specify any particular GPU models, CPU models, or other detailed hardware specifications used for experiments.
Software Dependencies No The paper mentions 'Py Torch CUDA tools' but does not specify version numbers for any software dependencies required to replicate the experiment.
Experiment Setup Yes The number of hierarchical layers H and the coarsening ratios for each step α1, . . . , αH are predefined as hyperparameters. ... In Table 3, l stands for the number of layers for Graph SAGE and Graphormer-SAMPLE, while number of Transformer layers at horizontal blocks for HSGT. d stands for the hidden dimension for all models. ... Here we study the impact of p value with experiments and summarize the results in Table 5...