GNN

GCN GCN originates from the article SEMI-SUPERVISED CLASSIFICATION WITH GRAPH CONVOLUTIONAL NETWORKS. Suppose there is a undirected graph G(V,E) with vertices V and edges E. It has an adjacency matrix A(VxV). Each node has a feature and the feature matrix is X(VxN) We want to classify the nodes into M categories using neural network f(X,A). $$ f(X,A) = softmax(\hat{A} ReLU(\hat{A} XW(0))W(1)) $$ where $$ \hat{A} = D^{-\frac{1}{2}}AD^{-\frac{1}{2}} $$ It computes the symmetrically normalized adjacency matrix of G....