Paper: Self-supervised Heterogeneous Graph Neural Network with Co-contrastive Learning
Code from author: https://github.com/liun-online/HeCo
Clone the Openhgnn-DGL
python main.py -m HeCo -d acm4HeCo -t node_classification -g 0 --use_best_configCandidate dataset: acm4HeCo
If you do not have gpu, set -gpu -1.
acm4HeCo
Node classification
| Node classification | acm4HeCo (Macro-F1 / Micro-F1 / AUC) |
|---|---|
| paper | 89.04 / 88.71 / 96.55 |
| OpenHGNN | 88.66 / 88.35 / 96.90 (mean of 10 random seeds) |
The model is trained in unsupervisied node classification.
hidden_dim = 64
max_epoch = 10000
eva_lr = 0.05
eva_wd = 0
patience = 5
learning_rate = 0.0008
weight_decay = 0
tau = 0.8
feat_drop = 0.3
attn_drop = 0.5
sample_rate = author-7_subject-1
lam = 0.5Best config can be found in best_config
Nian Liu, Tianyu Zhao[GAMMA LAB]
Submit an issue or email to [email protected].