Paper:Unsupervised Attributed Multiplex Network Embedding
Code from author:https://github.com/pcy1302/DMGI
Clone the Openhgnn-DGL
python main.py -m DMGI -t node_classification -d acm_han_raw -g 0 --use_best_configCandidate dataset: acm_han_raw
If you do not have gpu, set -gpu -1.
acm_han_raw/imdb4GTN
NOTE: DMGI can handle imdb dataset, we will add the dataset in our further work.
Node classification
| Node classification | acm | imdb4GTN |
|---|---|---|
| paper | 89.8 | --- |
| OpenHGNN | 89.73 | 52.52 |
The model is trained in unsupervisied node classification.
learning_rate = 0.0005
l2_coef = 0.0001
sc = 3
dropout = 0.5
reg_coef = 0.001
sup_coef = 0.1
patience =20
hid_unit = 64
num_heads = 1
max_epoch = 10000
isSemi = False
isBias = False
isAttn = FalseBest config can be found in best_config
Siyuan Zhang, Tianyu Zhao[GAMMA LAB]
Submit an issue or email to 1580124318@qq.com.