Neural Relational Inference for Interacting Systems

TL;DR

This paper introduces Neural Relational Inference (NRI), an unsupervised model that learns interaction graphs and dynamics from observational data using a variational auto-encoder with graph neural networks.

Contribution

The paper presents a novel unsupervised approach to infer interaction structures and dynamics in complex systems using neural networks and graph-based models.

Findings

Accurately recovers ground-truth interactions in simulated systems

Finds interpretable structures in real motion and sports data

Predicts complex dynamics effectively

Abstract

Interacting systems are prevalent in nature, from dynamical systems in physics to complex societal dynamics. The interplay of components can give rise to complex behavior, which can often be explained using a simple model of the system's constituent parts. In this work, we introduce the neural relational inference (NRI) model: an unsupervised model that learns to infer interactions while simultaneously learning the dynamics purely from observational data. Our model takes the form of a variational auto-encoder, in which the latent code represents the underlying interaction graph and the reconstruction is based on graph neural networks. In experiments on simulated physical systems, we show that our NRI model can accurately recover ground-truth interactions in an unsupervised manner. We further demonstrate that we can find an interpretable structure and predict complex dynamics in real motion capture and sports tracking data.