zation-of-Graph-Convolutional-Networks-in-PyTorch 联合开发网

Pudn.com > 下载中心 > 数据可视化 > zation-of-Graph-Convolutional-Networks-in-PyTorch

zation-of-Graph-Convolutional-Networks-in-PyTorch

所属分类：数据可视化
开发工具：Python
文件大小：21432KB
下载次数：0
上传日期：2020-03-14 02:32:50
上传者：sh-1993

说明：基于PyTorch中的图卷积网络，利用t-SNE算法增加了测试集结果的可视化。
(Based on Graph Convolutional Networks in PyTorch, visualization of test set results was added with t-SNE algorithm.)

文件列表:

.idea (0, 2020-03-14)
.idea\PyGCN_Visualization.iml (453, 2020-03-14)
.idea\encodings.xml (135, 2020-03-14)
.idea\misc.xml (319, 2020-03-14)
.idea\modules.xml (290, 2020-03-14)
.idea\workspace.xml (26456, 2020-03-14)
2D.png (135594, 2020-03-14)
3D.gif (21638536, 2020-03-14)
LICENCE (1071, 2020-03-14)
data (0, 2020-03-14)
data\cora (0, 2020-03-14)
data\cora\cora.cites (69928, 2020-03-14)
data\cora\cora.content (7823427, 2020-03-14)
figure.png (49982, 2020-03-14)
pygcn (0, 2020-03-14)
pygcn\MNIST_with_t-SNE.py (1506, 2020-03-14)
pygcn\S_data_with_t-SNE.py (1623, 2020-03-14)
pygcn\__init__.py (135, 2020-03-14)
pygcn\demos.py (239, 2020-03-14)
pygcn\layers.py (1806, 2020-03-14)
pygcn\models.py (751, 2020-03-14)
pygcn\test.py (2929, 2020-03-14)
pygcn\train.py (7180, 2020-03-14)
pygcn\try_of_BuidingGraph.py (1805, 2020-03-14)
pygcn\try_of_LoadData.py (2517, 2020-03-14)
pygcn\try_of_normalize.py (1950, 2020-03-14)
pygcn\try_of_onehot.py (2102, 2020-03-14)
pygcn\utils.py (5171, 2020-03-14)
setup.py (553, 2020-03-14)

Modification of Graph Convolutional Networks in PyTorch ==== ## introduction **Modification of [Graph Convolutional Networks in PyTorch](https://github.com/tkipf/pygcn), visualization of test set results was added in this version with t-SNE algorithm.** You can cancel the visualization of ground truth and test set results with `--no_visual` for quick debugging. Thanks to the original author [Thomas Kipf](https://github.com/tkipf) for the open source code, which provided me with the opportunity to learn. The visualization results are shown below: ![visualization in 2-dim](2D.png) ![visualization in 3-dim](3D.gif) In addition, the division of the training set in this project is not the same as the division method in the paper. In the paper, 4 samples are selected from each type of sample to form the training set, while 100 samples are randomly selected as the training set in the project. PyTorch implementation of Graph Convolutional Networks (GCNs) for semi-supervised classification [1]. For a high-level introduction to GCNs, see: Thomas Kipf, [Graph Convolutional Networks](http://tkipf.github.io/graph-convolutional-networks/) (2016) ![Graph Convolutional Networks](figure.png) Note: There are subtle differences between the TensorFlow implementation in https://github.com/tkipf/gcn and this PyTorch re-implementation. This re-implementation serves as a proof of concept and is not intended for reproduction of the results reported in [1]. This implementation makes use of the Cora dataset from [2]. ## Installation ```python setup.py install``` ## Requirements * PyTorch 0.4 or 0.5 * Python 2.7 or 3.6 ## Usage ```python train.py``` The following optional hyperparameters can be set before running: - no_cuda:Disables CUDA training. - fastmode:Validate during training pass. - seed:Random seed. - epochs:Number of epochs to train. - lr:Initial learning rate. - weight_decay:Weight decay (L2 loss on parameters). - hidden:Number of hidden units. - dropout:Dropout rate (1 - keep probability). ## References [1] [Kipf & Welling, Semi-Supervised Classification with Graph Convolutional Networks, 2016](https://arxiv.org/abs/1609.02907) [2][Sen et al., Collective Classification in Network Data, AI Magazine 2008](http://linqs.cs.umd.edu/projects/projects/lbc/)

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：