# 图嵌入和图神经网络
> 转载自：https://leovan.me/cn/2020/04/graph-embedding-and-gnn/
## 图嵌入
### Random Walk
### Matrix Fractorization
### Meta Paths
### Deep Learning
### Others
## 图神经网络
### Graph Neural Networks
### Graph Convolutional Networks
GCN 方法可以分为两大类：基于频谱（Spectral Methods）和基于空间（Spatial Methods）的方法。
### Graph Recurrent Networks
### Graph Attention Networks
GAT 中的注意力架构有如下几个特点：
1. 针对节点对的计算是并行的，因此计算过程是高效的。
2. 可以处理不同度的节点并对邻居分配对应的权重。
3. 可以容易地应用到归纳学习问题中去。
### 应用
图神经网络已经被应用在监督、半监督、无监督和强化学习等多个领域。下图列举了 GNN 在不同领域内相关问题中的应用，具体模型论文请参考 Graph Neural Networks: A Review of Methods and Applications 原文。
### 开源实现
<table>
<thead>
<tr>
<th>项目</th>
<th>框架</th>
</tr>
</thead>
<tbody>
<tr>
<td><a href="https://github.com/rusty1s/pytorch_geometric" target="_blank">rusty1s/pytorch_geometric</a></td>
<td><i class="icon icon-pytorch">PyTorch</i></td>
</tr>
<tr>
<td><a href="https://github.com/dmlc/dgl" target="_blank">dmlc/dgl</a></td>
<td><i class="icon icon-pytorch">PyTorch</i>, <i class="icon icon-tensorflow">TF</i> & <i class="icon icon-mxnet">MXNet</i></td>
</tr>
<tr>
<td><a href="https://github.com/alibaba/euler" target="_blank">alibaba/euler</a></td>
<td><i class="icon icon-tensorflow">TF</i></td>
</tr>
<tr>
<td><a href="https://github.com/alibaba/graph-learn" target="_blank">alibaba/graph-learn</a></td>
<td><i class="icon icon-tensorflow">TF</i></td>
</tr>
<tr>
<td><a href="https://github.com/deepmind/graph_nets" target="_blank">deepmind/graph_nets</a></td>
<td><i class="icon icon-tensorflow">TF</i> & <i class="icon icon-sonnet">Sonnet</i></td>
</tr>
<tr>
<td><a href="https://github.com/facebookresearch/PyTorch-BigGraph" target="_blank">facebookresearch/PyTorch-BigGraph</a></td>
<td><i class="icon icon-pytorch">PyTorch</i></td>
</tr>
<tr>
<td><a href="https://github.com/tencent/plato" target="_blank">tencent/plato</a></td>
<td></td>
</tr>
<tr>
<td><a href="https://github.com/PaddlePaddle/PGL" target="_blank">PaddlePaddle/PGL</a></td>
<td><i class="icon icon-paddlepaddle"></i> PaddlePaddle</td>
</tr>
<tr>
<td><a href="https://github.com/Accenture/AmpliGraph" target="_blank">Accenture/AmpliGraph</a></td>
<td><i class="icon icon-tensorflow">TF</i></td>
</tr>
<tr>
<td><a href="https://github.com/danielegrattarola/spektral" target="_blank">danielegrattarola/spektral</a></td>
<td><i class="icon icon-tensorflow">TF</i></td>
</tr>
<tr>
<td><a href="https://github.com/THUDM/cogdl/" target="_blank">THUDM/cogdl</a></td>
<td><i class="icon icon-pytorch">PyTorch</i></td>
</tr>
<tr>
<td><a href="https://github.com/DeepGraphLearning/graphvite" target="_blank">DeepGraphLearning/graphvite</a></td>
<td><i class="icon icon-pytorch">PyTorch</i></td>
</tr>
</tbody>
</table>
![]()