Interaction Networks for Learning about Objects, Relations and Physics

TL;DR

The paper introduces interaction networks, a neural network-based model capable of reasoning about objects, their relations, and physics to perform accurate simulations, inferences, and generalizations across complex physical systems.

Contribution

It presents the first general-purpose, learnable physics engine that models object interactions and physics using deep neural networks, enabling flexible reasoning in diverse domains.

Findings

Accurately simulates physical trajectories of multiple objects.

Estimates abstract quantities like energy.

Generalizes to systems with different objects and configurations.

Abstract

Reasoning about objects, relations, and physics is central to human intelligence, and a key goal of artificial intelligence. Here we introduce the interaction network, a model which can reason about how objects in complex systems interact, supporting dynamical predictions, as well as inferences about the abstract properties of the system. Our model takes graphs as input, performs object- and relation-centric reasoning in a way that is analogous to a simulation, and is implemented using deep neural networks. We evaluate its ability to reason about several challenging physical domains: n-body problems, rigid-body collision, and non-rigid dynamics. Our results show it can be trained to accurately simulate the physical trajectories of dozens of objects over thousands of time steps, estimate abstract quantities such as energy, and generalize automatically to systems with different numbers and configurations of objects and relations. Our interaction network implementation is the first general-purpose, learnable physics engine, and a powerful general framework for reasoning about object and relations in a wide variety of complex real-world domains.