GCN layers can be written as graph convolutions

[[concept]]

Idea

We can write a graph convolutional network layer as a graph convolution. This is easy to see by simply writing $x_{\ell }=\sigma (S{x}_{\ell -1}{H}_{\ell })$ Which is exactly the form of a graph convolution with

• $K=2$
• $S=D_{\text{binary}}^{-1/2}{A}_{\text{binary}}{D}_{\text{binary}}^{-1/2}$
• $H_{\ell ,0}=0,H_{\ell ,1}=H_{\ell }$

And $A_{\text{binary}}$ is the binary adjacency matrix (given as A[A>0] in Python)

Mentions

TABLE
FROM [[]]
 
FLATTEN choice(contains(artist, this.file.link), 1, "") + choice(contains(author, this.file.link), 1, "") + choice(contains(director, this.file.link), 1, "") + choice(contains(source, this.file.link), 1, "") as direct_source
 
WHERE !direct_source