chengaf / afcheng.github.io

blog
https://chengaf.github.io/afcheng.github.io/
0 stars 0 forks source link

GCN原理简略思路 | Little's Blogs #2

Open chengaf opened 4 years ago

chengaf commented 4 years ago

https://chengaf.github.io/afcheng.github.io/2019/10/30/GCN%E5%8E%9F%E7%90%86%E7%AE%80%E7%95%A5%E6%80%9D%E8%B7%AF/

写在前面,GCN提出目的、输入输出、方法分类。 GCN提出目的:在Graph这一广泛使用的数据结构上做深度学习。传统CNN无法直接迁移到Graph上,因此提出GCN这一研究体系/方法框架。 GCN方法分类:时域、频域(有理论保证是目前的主流方法,这里的总结)。 输入:Graph的邻接矩阵 A\in R^{N\times N} ,节点的特征矩阵 X\in R^{N\times F} 。 输出:节

NSF20 commented 4 years ago

腻害了哇