la 4f8ffa2b21 | 2 years ago | |
---|---|---|
.. | ||
README.md | 2 years ago |
Paper:Scalable Graph Neural Networks for Heterogeneous Graphs
Code from author:https://github.com/facebookresearch/NARS
Clone the Openhgnn-DGL
python main.py -m NARS -t node_classification -d acm4NARS -g 0 --use_best_config
Candidate dataset: acm4NARS
If you do not have gpu, set -gpu -1.
acm4NARS
NOTE: NARS can handle mag oag dataset, we will add these two datasets in our further work.
Number of nodes
paper | 4025 |
---|---|
author | 17431 |
field | 73 |
Number of edges
paper-author | 13407 |
---|---|
paper-field | 4025 |
Subsets: paper-author, paper-field
Node classification
accuracy | |
---|---|
acm4NARS | 0.93 |
NARS
NARS is composed of WeightedAggregator and SIGN.
WeightedAggregator
Get new features by multiplying the old features by the weight matrix.
SIGN
The MLP classifier. It is composed of a several linear layers. Then project the node embeddings to the vector space to predict the type of the nodes.
R = 2
input_dropout = True
cpu_preprocess = True
ff_layer = 2
Best config can be found in best_config
Tianyu Zhao, Yibo Li[GAMMA LAB]
Submit an issue or email to liushiliushi0@gmail.com.
OpenHGNN是由北邮GAMMA Lab开发的基于PyTorch和DGL的开源异质图神经网络工具包。
Python Markdown Shell
Dear OpenI User
Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.
For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》