Relational VAE: A Continuous Latent Variable Model for Graph Structured Data

TL;DR

This paper introduces a novel graph network-based variational autoencoder that models complex graph-structured data probabilistically, enabling applications in wind farm monitoring and meta-learning with improved flexibility.

Contribution

It extends graph networks with probabilistic modeling using Variational Bayes, incorporating graph-structured latent variables for enhanced relational data analysis.

Findings

Effective structured density modeling for wind farm data

Successful meta-learning on Gaussian Process data

Neural Processes can be interpreted within this framework

Abstract

Graph Networks (GNs) enable the fusion of prior knowledge and relational reasoning with flexible function approximations. In this work, a general GN-based model is proposed which takes full advantage of the relational modeling capabilities of GNs and extends these to probabilistic modeling with Variational Bayes (VB). To that end, we combine complementary pre-existing approaches on VB for graph data and propose an approach that relies on graph-structured latent and conditioning variables. It is demonstrated that Neural Processes can also be viewed through the lens of the proposed model. We show applications on the problem of structured probability density modeling for simulated and real wind farm monitoring data, as well as on the meta-learning of simulated Gaussian Process data. We release the source code, along with the simulated datasets.