BriefGPT.xyz
Nov, 2020
可以最终模拟本地结构的图卷积
Graph convolutions that can finally model local structure
HTML
PDF
Rémy Brossard, Oriel Frigo, David Dehaene
TL;DR
我们提出了一种很简单的方法来修改GIN卷积,以便使网络可以检测到很小的环,这对于在化学上下文中对图形子结构进行分析的下游任务非常重要。在真实的分子属性数据集上进行测试,我们的模型始终改善了大型多任务数据集的性能,无论是在全局还是在每个任务的设置中。
Abstract
Despite quick progress in the last few years, recent studies have shown that modern
graph neural networks
can still fail at very simple tasks, like detecting
small cycles
. This hints at the fact that current netw
→