BriefGPT.xyz
Feb, 2024
通过无监督学习在图上学习MLP以加速图推理
Graph Inference Acceleration by Learning MLPs on Graphs without Supervision
HTML
PDF
Zehong Wang, Zheyuan Zhang, Chuxu Zhang, Yanfang Ye
TL;DR
通过无监督学习简化后的多层感知器(MLPs)在图表上进行学习,以增强泛化效果,特别是在未见节点的设置中,实现了显著的性能提升(7-26%)和图表推断加速(90-126倍),在大规模图表数据集上优于现有基准方案。
Abstract
graph neural networks
(GNNs) have demonstrated effectiveness in various graph learning tasks, yet their reliance on
message-passing constraints
their deployment in latency-sensitive applications such as
→