This is the code of paper: SiMilarity-Enhanced Homophily for Multi-View Heterophilous Graph Clustering
- 'requirements.txt'
Datasets are included in ./data/.
| Dataset | #Clusters | #Nodes | #Features | Graphs | HR |
|---|---|---|---|---|---|
| ACM | 3 | 3025 | 1830 1830 |
|
0.82 0.64 |
| DBLP | 4 | 4057 | 334 334 334 |
|
0.80 0.67 0.32 |
| Texas | 5 | 183 | 1703 1703 |
|
0.09 0.09 |
| Chameleon | 5 | 22777 | 2325 2325 |
|
0.23 0.23 |
# Test SMHGC on ACM dataset
python SMHGC.py --dataset 'acm' --train 0 --use_cuda True --cuda_device 0
# Test SMHGC on DBLP dataset
python SMHGC.py --dataset 'dblp' --train 0 --use_cuda True --cuda_device 0
# Test SMHGC on Texas dataset
python SMHGC.py --dataset 'texas' --train 0 --use_cuda True --cuda_device 0
# Test SMHGC on Chameleon dataset
python SMHGC.py --dataset 'chameleon' --train 0 --use_cuda True --cuda_device 0# Train SMHGC on ACM dataset
python SMHGC.py --dataset 'acm' --train 1 --use_cuda True --cuda_device 0
# Train SMHGC on DBLP dataset
python SMHGC.py --dataset 'dblp' --train 1 --use_cuda True --cuda_device 0
# Train SMHGC on Texas dataset
python SMHGC.py --dataset 'texas' --train 1 --use_cuda True --cuda_device 0
# Train SMHGC on Chameleon dataset
python SMHGC.py --dataset 'chameleon' --train 1 --use_cuda True --cuda_device 0Parameters: More parameters and descriptions can be found in the script and paper.
| NMI% | ARI% | ACC% | F1% | |
|---|---|---|---|---|
| ACM | 81.1 |
83.2 |
93.9 |
93.9 |
| DBLP | 76.2 |
81.9 |
92.4 |
91.8 |
| Texas | 41.8 |
46.9 |
71.3 |
49.8 |
| Chameleon | 20.0 |
15.1 |
42.1 |
41.3 |