Adversarial Stein Training for Graph Energy Models

TL;DR

This paper introduces an adversarial Stein training method for graph energy models using GNNs, enabling effective graph generation by minimizing Stein discrepancy and employing Langevin dynamics for sampling.

Contribution

The work presents a novel adversarial Stein discrepancy approach for training energy-based models on graphs, improving graph generation performance.

Findings

Achieves competitive graph generation results

Uses Langevin dynamics for sampling

Introduces a new training method for graph EBMs

Abstract

Learning distributions over graph-structured data is a challenging task with many applications in biology and chemistry. In this work we use an energy-based model (EBM) based on multi-channel graph neural networks (GNN) to learn permutation invariant unnormalized density functions on graphs. Unlike standard EBM training methods our approach is to learn the model via minimizing adversarial stein discrepancy. Samples from the model can be obtained via Langevin dynamics based MCMC. We find that this approach achieves competitive results on graph generation compared to benchmark models.