Discovering physical concepts with neural networks

TL;DR

This paper introduces a neural network architecture inspired by human physical reasoning, enabling machine-assisted scientific discovery from data without prior assumptions, demonstrated through toy examples that reveal physical parameters and concepts.

Contribution

The authors develop a neural network model that mimics human reasoning to discover physical concepts directly from data, advancing scientific discovery methods.

Findings

Network identifies physically relevant parameters

Exploits conservation laws for predictions

Helps gain conceptual insights like heliocentrism

Abstract

Despite the success of neural networks at solving concrete physics problems, their use as a general-purpose tool for scientific discovery is still in its infancy. Here, we approach this problem by modelling a neural network architecture after the human physical reasoning process, which has similarities to representation learning. This allows us to make progress towards the long-term goal of machine-assisted scientific discovery from experimental data without making prior assumptions about the system. We apply this method to toy examples and show that the network finds the physically relevant parameters, exploits conservation laws to make predictions, and can help to gain conceptual insights, e.g. Copernicus' conclusion that the solar system is heliocentric.