Simplifying Graph Convolutional Networks · CiteArk

Simplifying Graph Convolutional Networks

公开

作者：Felix Wu、Tianyi Zhang、Amauri Holanda de Souza Jr.、Christopher Fifty、Tao Yu、Kilian Q. Weinberger

ICML 2019 arXiv 2019 机器学习 (cs.LG)机器学习 (stat.ML)

研究摘要

基于论文

This paper derives Simple Graph Convolution (SGC) by removing intermediate nonlinearities from a graph convolutional network and collapsing its weight matrices into one linear classifier. The resulting method applies a fixed, parameter-free graph propagation filter to node features before multinomial logistic regression. A spectral analysis relates propagation with self-loops to low-pass filtering and proves that self-loops shrink the normalized-Laplacian spectrum. Experiments compare SGC with graph neural-network baselines on citation and social networks and adapt it to text classification, geolocation, relation extraction, zero-shot image classification, graph classification, and molecular prediction, emphasizing accuracy, training time, stability, and known failure cases.

这项研究意味着什么

研究解读

The work shows that repeated graph smoothing, rather than nonlinear transformations at every layer, can explain much of GCN performance on several benchmarks. SGC is therefore a useful simple and fast baseline. The paper does not claim universal equivalence: it reports a small loss on MR, substantially worse QM8 error than more expressive graph models, and graph-classification performance far behind GIN.

SGC 复现：两处贴近，一处偏离

对 SGC 论文中三个公开引文网络结果的十次运行复现显示：Cora 与 Pubmed 在数值上贴近论文，Citeseer 则取决于存在来源冲突的训练轮数。主比较遵循论文规定的 100 轮；README 的 150 轮仅作为敏感性检查。

复现差距与轮数分歧

十次运行的均值相对论文值：Cora 高 0.05 个百分点，Pubmed 高 0.06 个百分点，Citeseer 在论文规定的 100 轮下低 0.70 个百分点。Citeseer 的 150 轮敏感性均值虽升至 71.85%，但该分支源于论文与 README 的条件冲突，不能按结果取代主比较；Pubmed 的数值证据仍因独立评审不可用而保持“不确定”。

查看数值与说明

相对各自论文值，十次运行均值的差为 Cora +0.05、Citeseer −0.70、Pubmed +0.06 个百分点。Citeseer 的 100 轮主分支十次运行均为 71.2%；150 轮敏感性分支介于 71.8% 与 71.9%，均值 71.85%，论文值为 71.9%。

Cora｜十次运行测试准确率均值: 81.05
Cora｜论文测试准确率: 81
Citeseer｜100 轮十次运行测试准确率均值: 71.2
Citeseer｜论文测试准确率: 71.9
Pubmed｜十次运行测试准确率均值: 78.96
Pubmed｜论文测试准确率: 78.9
Citeseer｜150 轮敏感性均值: 71.85

扩展与历史报告

扩展视图

仓库信息