最近更新时间:2022.12.25
最早更新时间:2022.9.27
本文统计各论文中常用的异质图数据集信息。
整体表格我做了个石墨文档,但是还没整理好,以后再公开发布吧。
任务:节点分类,预测paper节点所属的venue(会议或期刊)(共有349类)
Leaderboard:https://ogb.stanford.edu/docs/leader_nodeprop/#ogbn-mag
可通过PyG直接加载。
任务:节点分类,预测venue或author所属的类别
数据来自Re31:读论文 metapath2vec: Scalable Representation Learning for Heterogeneous Networks一文,最初出自ArnetMiner: Extraction and Mining of Academic Social Networks一文。
可通过PyG直接加载。
数据来自https://www.aminer.org/aminernetwork(论文也是ArnetMiner: Extraction and Mining of Academic Social Networks),在A multilayered approach for link prediction in heterogeneous complex networks一文中被用作链路预测任务(但是在这篇文章里叫DBLP,是不是很无语)。
任务:计算节点相似度
数据来自Re31:读论文 metapath2vec: Scalable Representation Learning for Heterogeneous Networks一文,最初出自Pathsim: Meta path-based top-k similarity search in heterogeneous information networks一文。
不便使用dropbox的读者如需下载数据,可在该GitHub项目README文件中展示的百度网盘链接里下载:https://github.com/PolarisRisingWar/HGNN_Collection
加载方式可参考我写的代码:https://github.com/PolarisRisingWar/HGNN_Collection/blob/master/load_data/dbis_pyg.py
任务:节点分类,预测author所属的research areas(共有4类)
数据来自MAGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding一文,出自Graph-based Consensus Maximization among Multiple Supervised and Unsupervised Models和Graph regularized transductive classification on heterogeneous information networks。
可通过PyG直接加载。
任务:节点分类,预测movie所属的category(共有3类)
数据来自MAGNN: Metapath Aggregated Graph Neural Network for Heterogeneous Graph Embedding一文。
可通过PyG直接加载。
(和下面的IMDB (Simple-HGN)用的是同一套原始数据集)
任务:节点分类,预测movie标签(共有5类)
数据来自Are We Really Making Much Progress? Revisiting, Benchmarking, and Refining Heterogeneous Graph Neural Networks
任务:链路预测(user-to-artist)
数据来自Are We Really Making Much Progress? Revisiting, Benchmarking, and Refining Heterogeneous Graph Neural Networks,出自HetRec 2011
可通过PyG直接加载。
(我在石墨文档中的统计信息是通过2022年9月下载的数据计算得到的)
原始数据是https://files.grouplens.org/datasets/movielens/ml-latest-small.zip。
可通过PyG直接加载。
任务:节点分类(paper,共3类)
数据来自Are We Really Making Much Progress? Revisiting, Benchmarking, and Refining Heterogeneous Graph Neural Networks,出自Heterogeneous graph attention network
任务:节点分类(book,共7类)
数据来自Are We Really Making Much Progress? Revisiting, Benchmarking, and Refining Heterogeneous Graph Neural Networks,出自Heterogeneous Network Representation Learning: A Unified Framework with Survey and Benchmark。
任务:链路预测
leaderboard:https://ogb.stanford.edu/docs/leader_linkprop/#ogbl-biokg