search

Yanqi-Chen / GWNN

Course project. A implementation of Graph Wavelet Neural Network (ICLR 2019)
GNU General Public License v3.0
10 stars 3 forks source link

readme

GWNN License

GNN课程大作业，对

Graph Wavelet Neural Network. Bingbing Xu, Huawei Shen, Qi Cao, Yunqi Qiu, Xueqi Cheng. ICLR, 2019. [Paper]

的复现，包含了Appendix D提到的利用Чебышёв多项式近似快速求小波基。

依赖

Python版本为3.7.5，主要包版本如下所示

cudatoolkit               10.1.243
numpy                     1.17.3
pytorch                   1.3.1
torchvision               0.4.2
scikit-learn              0.21.3
scipy                     1.3.1
tensorboard               2.0.0
tensorflow                2.0.0
tensorflow-base           2.0.0

参数选项

输入参数

  --dataset     STR Which dataset to use.           'cora', 'citeseer' or 'pubmed'. Default is 'cora'.
  --save-path       STR Target directory for saving models. Default is './models'

模型参数

  --epochs      INT Number of training epochs.      Default is 200.
  --hidden      INT Number of units in hidden layer.    Default is 16.
  --weight-decay    FLOAT   Adam weight decay.          Default is 5e-4.
  --learning-rate   FLOAT   Learning rate.              Default is 0.01.
  --dropout     FLOAT   Dropout probability.            Default is 0.5.
  --approximation-order INT Chebyshev polynomial order.     Default is 3.
  --threshold       FLOAT   Sparsification parameter.       Default is 1e-4.
  --scale       FLOAT   Scaling parameter.          Default is 1.0.
  --fast        BOOL    Use fast graph wavelets with Chebyshev polynomial approximation.

例子

python train.py --dataset cora --fast --approximation-order 3