Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
Guangyu Zhou eaa51ac4dc | 3 months ago | |
---|---|---|
.. | ||
gcnii_trainer.py | 3 months ago | |
readme.md | 1 year ago |
Note that our implementation is little different with the author's in the optimizer. The author applied different weight decay coefficient on learnable paramenters, while TensorLayerX has not support this feature.
Dataset | # Nodes | # Edges | # Classes |
---|---|---|---|
Cora | 2,708 | 10,556 | 7 |
Citeseer | 3,327 | 9,228 | 6 |
Pubmed | 19,717 | 88,651 | 3 |
Refer to Planetoid. |
TL_BACKEND="tensorflow" python gcnii_trainer.py --dataset cora --lr 0.01 --num_layers 64 --alpha 0.1 --hidden_dim 64 --lambd 0.5 --drop_rate 0.3 --l2_coef 0.001
TL_BACKEND="tensorflow" python gcnii_trainer.py --dataset citeseer --lr 0.01 --num_layers 32 --alpha 0.1 --hidden_dim 256 --lambd 0.5 --drop_rate 0.3 --l2_coef 0.001
TL_BACKEND="tensorflow" python gcnii_trainer.py --dataset pubmed --lr 0.01 --num_layers 16 --alpha 0.1 --hidden_dim 256 --lambd 0.4 --drop_rate 0.3 --l2_coef 0.001
TL_BACKEND="paddle" python gcnii_trainer.py --dataset cora --lr 0.01 --num_layers 64 --alpha 0.1 --hidden_dim 64 --lambd 0.5 --drop_rate 0.3 --l2_coef 0.001
TL_BACKEND="paddle" python gcnii_trainer.py --dataset citeseer --lr 0.01 --num_layers 32 --alpha 0.1 --hidden_dim 256 --lambd 0.4 --drop_rate 0.4 --l2_coef 0.001
TL_BACKEND="paddle" python gcnii_trainer.py --dataset pubmed --lr 0.01 --num_layers 16 --alpha 0.1 --hidden_dim 256 --lambd 0.5 --drop_rate 0.7 --l2_coef 0.001
TL_BACKEND="torch" python gcnii_trainer.py --dataset cora --lr --lr 0.01 --num_layers 64 --alpha 0.1 --hidden_dim 64 --lambd 0.5 --drop_rate 0.3 --l2_coef 0.001
TL_BACKEND="torch" python gcnii_trainer.py --dataset citeseer --lr 0.01 --num_layers 64 --alpha 0.1 --hidden_dim 64 --lambd 0.6 --drop_rate 0.4 --l2_coef 0.001
TL_BACKEND="torch" python gcnii_trainer.py --dataset pubmed --lr 0.01 --num_layers 64 --alpha 0.1 --hidden_dim 64 --lambd 0.4 --drop_rate 0.6 --l2_coef 0.001
Dataset | Paper | Our(pd) | Our(tf) | Our(tf) |
---|---|---|---|---|
cora | 85.5 | 83.12(±0.47) | 83.23(±0.76) | 83.1(±0.9) |
pubmed | 73.4 | 72.04(±0.91) | 71.9(±0.7) | 71.4(±0.6) |
citeseer | 80.3 | 80.36(±0.65) | 80.1(±0.5) | 80.5(±0.3) |
Notice that we do not use the same regularization method as the paper do, as TensorlayerX currently do not support it.
No Description
Python C++ Cuda Markdown Text
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》