## Abstract 摘要

We present a scalable approach for semi-supervised learning on graph-structureddata that is based on an efficient variant of convolutional neural networks whichoperate directly on graphs. We motivate the choice of our convolutional archi-tecture via a localized first-order approximation of spectral graph convolutions.Our model scales linearly in the number of graph edges and learns hidden layerrepresentations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph datasetwe demonstrate that our approach outperforms related methods by a significantmargin.

## Notes 笔记

• 谱图卷积的一阶近似、逐层传播规则->一种图卷积神经网络；
• 这种模型如何运用在图节点的半监督分类任务上。

$g_\theta$可以看作是关于特征值矩阵$\Lambda$的函数，即:

$g_\theta(\Lambda)$可以用Chebyshev多项式$T_k(x)$的前k阶来拟合，即:

## Conclusion 结论

We have introduced a novel approach for semi-supervised classification on graph-structured data.Our GCN model uses an efficient layer-wise propagation rule that is based on a first-order approx-imation of spectral convolutions on graphs. Experiments on a number of network datasets suggestthat the proposed GCN model is capable of encoding both graph structure and node features in away useful for semi-supervised classification. In this setting, our model outperforms several recentlyproposed methods by a significant margin, while being computationally efficient.