Adversarial Stein Training for Graph Energy Models

Abstract

Learning distributions over graph-structured data is a challenging task. In this work we present an energy-based model (EBM) using graph neural networks (GNN) to learn permutation invariant unnormalized density functions on graphs. The model is trained via minimizing adversarial stein discrepancy. Samples from the model can be obtained via Langevin dynamics. We find that this approach achieves better results on graph generation compared to benchmark models.

Tasks

Reproductions