An Attention-Based Graph Neural Network for Heterogeneous Structural Learning

Authors: Huiting Hong, Hantao Guo, Yucheng Lin, Xiaoqing Yang, Zang Li, Jieping Ye4132-4139

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The experiments conducted on three public datasets demonstrate that our proposed models achieve significant and consistent improvements compared to state-of-the-art solutions.
Researcher Affiliation Collaboration 1AI Labs, Didi Chuxing, Beijing, China, 2Peking University, Beijing, China
Pseudocode No The paper describes the model architecture and equations but does not include an explicit pseudocode block or algorithm figure.
Open Source Code Yes Available at https://github.com/didi/hetsann
Open Datasets Yes We collected a movie graph from IMDB site3 and constructed two academic networks from DBLP (Ji et al. 2010) and AMiner (Tang et al. 2008) datasets respectively. ... 3https://www.imdb.com
Dataset Splits Yes The whole labeled dataset is randomly split into training set, validation set and test set by a ratio of 0.8:0.1:0.1.
Hardware Specification No The paper does not provide specific details on the hardware used, such as CPU or GPU models, or memory specifications.
Software Dependencies No The paper mentions "Adam solver (Kingma and Ba 2015)" but does not provide specific version numbers for this or any other software or libraries used.
Experiment Setup Yes All variations employ 3-layer Het SANN and each TAL consists of 8 attention heads. The output dimensions of each attention head are consistent to 8. The parameters are optimized via Adam solver (Kingma and Ba 2015) with a learning rate 0.001 for IMDB and 0.005 for other datasets. A regularization weight 0.0005 is applied to all trainable parameters. A dropout rate 0.6 (Srivastava et al. 2014) is implanted between hidden layers to stabilize our model training procedure. For the variant of Het SANN with suffix .V , the weight coefficients β1 =10 3 and β2 =10 5.