Graph Neural Networks | Deep Learning

Live Engine

Select Topic

easyGraph Neural Networks

A graph G = (V, E) has 5 nodes and adjacency matrix A. A GCN (Graph Convolutional Network) updates node features as H^{(l+1)} = σ(D^{-1/2} Ã D^{-1/2} H^{(l)} W^{(l)}), where Ã = A + I (self-loops added). What is the role of the D^{-1/2} Ã D^{-1/2} normalization?

Live Engine

Select Topic

easyGraph Neural Networks

A graph G = (V, E) has 5 nodes and adjacency matrix A. A GCN (Graph Convolutional Network) updates node features as H^{(l+1)} = σ(D^{-1/2} Ã D^{-1/2} H^{(l)} W^{(l)}), where Ã = A + I (self-loops added). What is the role of the D^{-1/2} Ã D^{-1/2} normalization?