Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
Theheavens 299b0523f9 | 2 years ago | |
---|---|---|
.. | ||
README.md | 2 years ago |
Clone the Openhgnn-DGL
python main.py -m MAGNN -t node_classification -d imdb4MAGNN -g 0
If you do not have gpu, set -gpu -1.
the dataset imdb4MAGNN/dblp4MAGNN is supported.
Node classification | Macro-F1 | Micro-F1 |
---|---|---|
20% | paper: 59.35% OpenHGNN: 61.52% | paper: 59.60% OpenHGNN: 61.45% |
40% | paper: 60.27% OpenHGNN: 62.65% | paper: 60.50% OpenHGNN: 62.59% |
60% | paper: 60.66% OpenHGNN: 63.37% | paper: 60.88% OpenHGNN: 63.26% |
80% | paper: 61.44% OpenHGNN: 63.86% | paper: 61.53% OpenHGNN: 63.81% |
Node classification | Macro-F1 | Micro-F1 |
---|---|---|
20% | paper: 59.35% OpenHGNN: 53.91% | paper: 59.60% OpenHGNN: 53.96% |
40% | paper: 60.27% OpenHGNN: 55.75% | paper: 60.50% OpenHGNN: 55.81% |
60% | paper: 60.66% OpenHGNN: 56.83% | paper: 60.88% OpenHGNN: 56.80% |
80% | paper: 61.44% OpenHGNN: 57.65% | paper: 61.53% OpenHGNN: 57.68% |
The perform of experiments are run in the setting of paper which use SVM classification, so there are a little bit different with semi-supervised node classification. And directly run the model is using semi-supervised node classification trainerflow.
Number of nodes
movie | 4278 |
---|---|
director | 2081 |
actor | 5257 |
Number of edges
movie-director | 4278 |
---|---|
movie-actor | 12828 |
Types of metapath: MDM, MAM, DMD, DMAMD, AMA, AMDMA
. Please note that the M
is movie, D
is director, A
is actor, and the edges above are all bidirectional.
Number of nodes
author | 4057 |
---|---|
paper | 14328 |
term | 7723 |
venue | 20 |
Number of edges
author-paper | 19645 |
---|---|
paper-term | 85810 |
paper-venue | 14328 |
Types of metapath: APA, APTPA, APVPA
. Please note that the A
is actor, P
is paper, T
is term, V
is venue, and the edges above are all bidirectional.
[TODO]
You can modify the parameters in openhgnn/config.ini
inter_attn_feats = 32 # the dimension of the attention vector in inter metapath
num_heads = 4 # the number of heads in the attention mechanism of intra metapath
num_layers = 2 # the number of layers.
Best config can be found in best_config
Jiahang Li[GAMMA LAB]
Submit an issue or email to ljh1064126026@bupt.edu.cn.
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》