[论文解读] Network In Graph Neural Network

Xiang Song, Runjie Ma+3 more|arXiv (Cornell University)|Nov 23, 2021

Advanced Graph Neural Networks参考文献 40被引用 6

一句话总结

该论文提出了一种模型无关的方法——图中网络神经网络（NGNN），通过在每个GNN层内插入非线性前馈网络来增强GNN的表征能力，避免过拟合和过平滑。NGNN在多种GNN模型和任务中均提升了性能，在OGB链接预测排行榜上取得了SOTA结果，其中在SEAL上hits@100指标提升达7.08%，在GraphSage+Edge-Attr上hits@20指标提升达6.22%。

ABSTRACT

Graph Neural Networks (GNNs) have shown success in learning from graph structured data containing node/edge feature information, with application to social networks, recommendation, fraud detection and knowledge graph reasoning. In this regard, various strategies have been proposed in the past to improve the expressiveness of GNNs. For example, one straightforward option is to simply increase the parameter size by either expanding the hid-den dimension or increasing the number of GNN layers. However, wider hidden layers can easily lead to overfitting, and incrementally adding more GNN layers can potentially result in over-smoothing.In this paper, we present a model-agnostic methodology, namely Network In Graph Neural Network (NGNN ), that allows arbitrary GNN models to increase their model capacity by making the model deeper. However, instead of adding or widening GNN layers, NGNN deepens a GNN model by inserting non-linear feedforward neural network layer(s) within each GNN layer. An analysis of NGNN as applied to a GraphSage base GNN on ogbn-products data demonstrate that it can keep the model stable against either node feature or graph structure perturbations. Furthermore, wide-ranging evaluation results on both node classification and link prediction tasks show that NGNN works reliably across diverse GNN architectures.For instance, it improves the test accuracy of GraphSage on the ogbn-products by 1.6% and improves the hits@100 score of SEAL on ogbl-ppa by 7.08% and the hits@20 score of GraphSage+Edge-Attr on ogbl-ppi by 6.22%. And at the time of this submission, it achieved two first places on the OGB link prediction leaderboard.

研究动机与目标

• 解决因隐藏维度或层数增加而导致深层GNN出现的过拟合和过平滑问题。
• 开发一种模型无关的方法，显著提升GNN的表征能力，而无需显著增加模型深度或宽度。
• 提升模型对节点特征和图结构扰动的鲁棒性。
• 在大规模图基准数据集上实现SOTA性能，同时计算开销极低。
• 支持多种训练方法，包括邻居采样、基于聚类的采样以及局部子图采样。

提出的方法

• 在每个GNN层内插入非线性前馈神经网络层，而非堆叠额外的GNN层。
• 独立地将NGNN应用于输入、隐藏或输出GNN层，以研究其位置对性能的影响。
• 在NGNN层中使用ReLU和Sigmoid等激活函数，引入额外的非线性。
• 将NGNN与多种GNN架构结合，包括GraphSage、GAT、GCN、AGDN和SEAL。
• 支持多种训练范式，包括全图训练、邻居采样、图聚类和局部子图采样。
• 通过在ogbn-products及其他基准上的消融实验，优化NGNN的放置位置和结构设计。

实验结果

研究问题

• RQ1 在GNN层内插入前馈网络是否能在不增加模型深度或宽度的前提下提升模型的表达能力？
• RQ2 NGNN在节点特征和图结构扰动下对模型鲁棒性有何影响？
• RQ3 NGNN在GNN架构中的哪个位置（输入、隐藏、输出）最有效？
• RQ4 NGNN能否在多种GNN模型和任务中实现SOTA性能？
• RQ5 与传统深层GNN方法相比，NGNN的内存和训练效率如何？

主要发现

• NGNN将GraphSage在ogbn-products上的测试准确率提升了1.6%，达到79.91%，当NGNN应用于隐藏层时。
• NGNN使SEAL在ogbl-ppa上的hits@100得分提升了7.08%，达到76.48%，采用全层NGNN配置。
• NGNN使GraphSage+Edge-Attr在ogbl-ppi上的hits@20得分提升了6.22%，达到83.45%，采用全层NGNN配置。
• NGNN结合AGDN+BoT+self-KD+C&S在ogbn-arxiv节点分类排行榜中获得第4名，且参数量更少。
• NGNN结合GAT+BoT在ogbn-proteins排行榜中获得第2名，展现出极高的效率与性能。
• NGNN在扰动下表现稳定，在ogbn-products数据集上对节点特征和图结构变化的性能均保持稳定。