We present a scalable approach for semi-supervised learning on graph-structured data that is based on an efficient variant of convolutional neural networks which operate directly on graphs. We motivate the choice of our convolutional architecture via a localized first-order approximation of spectral graph convolutions. Our model scales linearly in the number of graph edges and learns hidden layer representations that encode both local graph structure and features of nodes. In a number of experiments on citation networks and on a knowledge graph dataset we demonstrate that our approach outperforms related methods by a significant margin.
核心贡献 · Key contributions
提出一种可扩展的图卷积网络（GCN），用于图上的半监督学习。 Introduces a scalable Graph Convolutional Network (GCN) for semi-supervised learning on graphs.
通过谱图卷积的局部一阶近似来论证 GCN 架构的合理性。 Motivates the GCN architecture via a localized first-order approximation of spectral graph convolutions.
实现与图边数成线性缩放，并编码局部结构和节点特征。 Achieves linear scaling in the number of graph edges and encodes both local structure and node features.
在引文网络和知识图谱数据集上优于相关方法。 Outperforms related methods on citation networks and knowledge graph datasets.
提出重归一化技巧以提高数值稳定性和性能。 Proposes a renormalization trick to improve numerical stability and performance.
在运行时间上比现有最优方法更高效。 Demonstrates efficiency in wall-clock time compared to state-of-the-art methods.
局限 · Limitations
全批量梯度下降导致内存需求与数据集大小成线性关系。 Full-batch gradient descent leads to memory requirements linear in dataset size.
不天然支持边特征或有向图。 Does not naturally support edge features or directed graphs.
假设局部性和自连接与邻居边的重要性相等。 Assumes locality and equal importance of self-connections vs. neighbor edges.
小批量 SGD 需要谨慎处理 K 阶邻域以实现精确推理。 Mini-batch SGD requires careful handling of K-order neighborhoods for exact inference.