想起近 20 前的一部动画《最终流放》，其世界分为三个阵营，其中一个阵营是调停者，只提供飞船的引擎。而其他两国天天开战，但战舰都要依赖于调停者的飞行引擎。看似两国战舰威猛咄咄逼人，但实际上却无本质差异。
这就是技术霸权吧
6 Matching Annotations
- Nov 2022
-
-
-
zhuanlan.zhihu.com zhuanlan.zhihu.com
-
a_input = torch.cat([h.repeat(1, N).view(N * N, -1), h.repeat(N, 1)], dim=1).view(N, -1, 2 * self.out_features)
$$h_i$$与$$h_j$$是如何体现的，这明明是一个h啊
-
-
arxiv.org arxiv.org
-
On the other hand, we have non-spectral approaches (Duvenaud et al., 2015; Atwood & Towsley,2016; Hamilton et al., 2017), which define convolutions directly on the graph, operating on groupsof spatially close neighbors. One of the challenges of these approaches is to define an operator whichworks with different sized neighborhoods and maintains the weight sharing property of CNNs
Non-spectral approaches design the convolutions directly on graph. One of the challenges of these approaches is to define an operator which works with different sized neighborhoods and maintains the weight sharing property of CNNs
-
Nevertheless, there is an increasing interest in generalizing convolutions to the graph domain. Ad-vances in this direction are often categorized as spectral approaches and non-spectral approaches.
The approaches in graph domain are divided in spectral approaches and non-spectral approaches.
-
Graph NeuralNetworks (GNNs) were introduced in Gori et al. (2005) and Scarselli et al. (2009) as a generalizationof recursive neural networks that can directly deal with a more general class of graphs, e.g. cyclic,directed and undirected graphs
Researchers used GNN to deal with directed and undirected graphs
-
Early work used recursive neural networks to process data represented in graphdomains as directed acyclic graph
Early work used RNN to process graph data
-