Tutorial on Variational Graph Auto-Encoders

The loss function for variational graph autoencoder is pretty much the same as before. The first part is the reconstruction loss between the input adjacency matrix and the reconstructed adjacency matrix. More specifically, it is the binary cross-entropy between the target (A) and output (A') logits. The second part is the KL-divergence between q (Z X, A) and p(Z), where p(Z) N(0,1). It measures how closely our q (Z X, A) matches to p(Z).