Latent variable GNN

This repository consists some models tentatively work for the combination of latent variable inference and graph neural networks.

Contents

• gcn-lp-filter consists some models for node classification.
• graph-classifier-dgl consists some models for graph pooling and classification, using some examples of DGL.
• graph-classifier-vi consists some models combining variational inference models (e.g. VGAE, Planar flow, Normalizing flow, Inverse autoregressive flow, etc.) to graph neural networks for graph pooling and classification, evaluated on molecular classification tasks, based on Graph U-Nets .