Graph network for learning bi-directional physics

TL;DR

This paper introduces a graph neural network that learns both forward and inverse particle physics models, achieving high accuracy and recovering complex probability distributions without problem-specific regularization.

Contribution

The work presents a novel end-to-end graph network that implicitly learns physics models and infers unknown properties, improving generalizability over traditional physics-informed neural networks.

Findings

Achieves at least an order of magnitude higher accuracy in next state prediction.

Successfully recovers multi-modal probability distributions of physical parameters.

Demonstrates effectiveness across diverse datasets.

Abstract

In this work, we propose an end-to-end graph network that learns forward and inverse models of particle-based physics using interpretable inductive biases. Physics-informed neural networks are often engineered to solve specific problems through problem-specific regularization and loss functions. Such explicit learning biases the network to learn data specific patterns and may require a change in the loss function or neural network architecture hereby limiting their generalizabiliy. Our graph network is implicitly biased by learning to solve several tasks, thereby sharing representations between tasks in order to learn the forward dynamics as well as infer the probability distribution of unknown particle specific properties. We evaluate our approach on one-step next state prediction tasks across diverse datasets. Our comparison against related data-driven physics learning approaches reveals that our model is able to predict the forward dynamics with at least an order of magnitude higher accuracy. We also show that our approach is able to recover multi-modal probability distributions of unknown physical parameters.