Reproduce “Discover physical concepts and equations with machine learning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bao-Bing Li, Yi Gu, Shao-Feng Wu
This research introduces a machine learning framework that combines Variational Autoencoders and Neural Ordinary Differential Equations to simultaneously discover physical concepts and their governing equations directly from observed data. The framework accurately extracted fundamental laws and their associated abstract concepts across diverse systems, including heliocentrism, Newtonian gravity, Schrödinger's wave mechanics, and Pauli's spin-magnetic formulation.
Key insights
- It is possible to relax strong prior assumptions about how latent physical variables evolve, relying only on the assumption that they follow differential equations.
- The framework can simulate human physical reasoning, moving from observations to concept formulation and equation building, without requiring explicit human guidance on physical laws.
- AI can uncover abstract, hidden physical principles (e.g., spin quantization) even when experimental data does not overtly display them, reducing reliance on clear observational patterns.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.