Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
Guangyu Zhou eaa51ac4dc | 3 months ago | |
---|---|---|
.. | ||
han_trainer.py | 3 months ago | |
readme.md | 1 year ago |
This is an implementation of HAN
for heterogeneous graphs.
python han_trainer.py
for reproducing HAN's work on IMDB.
Note: this scripts only support
IMDB
, which means commandpython han_trainer.py --dataset ACM
will not run onACM
.
If you want to test the performance of other datasets, you are suggested to make some modification of the trainer script.
Reference performance numbers for the IMDB dataset:
(0.01, 200, 0.0001, 8, 0.8, 0.58178, 0.002811689883326394)
train test val = 400, 3478, 400, about 9% for trianing
Dataset | Paper(80% training) | Paper(60% training) | Paper(40% training) | Paper(20% training) | Our(tf) | Our(th) | Our(pd) |
---|---|---|---|---|---|---|---|
IMDB | 58.51 | 58.32 | 57.97 | 55.73 | 57.78(±0.51) | 55.66(±1.05) | 56.58(±0.51) |
TL_BACKEND="tensorflow" python3 han_trainer.py --n_epoch 200 --lr 0.01 --l2_coef 0.0001 --heads 8 --drop_rate 0.8
TL_BACKEND="torch" python3 han_trainer.py --n_epoch 200 --lr 0.01 --l2_coef 0.0001 --heads 16 --drop_rate 0.4
TL_BACKEND="paddle" python3 han_trainer.py --n_epoch 200 --lr 0.01 --l2_coef 0.0001 --heads 16 --drop_rate 0.4
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》