huawei-noah / Efficient-AI-Backbones

Efficient AI Backbones including GhostNet, TNT and MLP, developed by Huawei Noah's Ark Lab.
4.05k stars 707 forks source link

VIG节点连边问题 #200

Closed worker121 closed 1 year ago

worker121 commented 1 year ago

您好! 您论文中所提及的对于VIG构建节点之后选择K个最近邻的邻居来建边。请问这个最近邻是怎么定义的? 我看您的代码中在class DenseDilated(nn.Module):中将计算获得的节点之间的pairwise distance直接使用edge_index = edge_index[:, :, :, ::self.dilation]进行切片处理,这样处理后的节点的k个邻居貌似就直接是每隔dilation个距离就取一个当邻居,请问这怎么体现最近邻的概念呢?

iamhankai commented 1 year ago

这个是dilated GCN的概念,可以参考论文sec3.4:https://openaccess.thecvf.com/content_ICCV_2019/papers/Li_DeepGCNs_Can_GCNs_Go_As_Deep_As_CNNs_ICCV_2019_paper.pdf

worker121 commented 1 year ago

哦哦,好的,谢谢您