Learning on Large-scale Text-attributed Graphs via Variational Inference

Authors: Jianan Zhao, Meng Qu, Chaozhuo Li, Hao Yan, Qian Liu, Rui Li, Xing Xie, Jian Tang

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

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments on multiple data sets demonstrate the efficiency and effectiveness of the proposed approach 1.
Researcher Affiliation Collaboration Jianan Zhao1,3 , Meng Qu1,3 , Chaozhuo Li2 , Hao Yan4, Qian Liu5, Rui Li6, Xing Xie2, Jian Tang1,7,8 1Mila Qu ebec AI Institute, 2Microsoft Research Asia, 3Universit e de Montr eal 4Central South University, 5Sea AI Lab, 6Dalian University of Technology 7HEC Montr eal, 8Canadian Institute for Advanced Research (CIFAR)
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes 1Codes are available at https://github.com/Andy JZhao/GLEM.
Open Datasets Yes Datasets. Three TAG node classification benchmarks are used in our experiment, including ogbnarxiv, ogbn-products, and ogbn-papers100M (Hu et al., 2020).
Dataset Splits Yes The statistics of these datasets are shown in Table 1. #Nodes #Edges Avg. Node Degree Train / Val / Test (%) ogbn-arxiv (Arxiv) 169,343 1,166,243 13.7 54 / 18 / 28 ogbn-products (Products) 2,449,029 61,859,140 50.5 8 / 2 / 90 ogbn-papers100M (Papers) 111,059,956 1,615,685,872 29.1 78 / 8 / 14
Hardware Specification Yes The maximum batch size (max bsz.) and time/epoch are tested on a single NVIDIA Tesla V100 32GB GPU.
Software Dependencies No The paper mentions using DeBERTa as the LM model but does not provide specific version numbers for any software dependencies like programming languages, libraries (e.g., PyTorch, TensorFlow), or other packages.
Experiment Setup Yes For fair comparison against other feature learning methods such as GIANT, the hyper-parameters of GNNs are set to the best settings described in the paper or in the official repository, other parameters are tuned by grid search.