/AI-Descartes

Open source package for accelerated symbolic discovery of fundamental laws.

Primary LanguageJavaMIT LicenseMIT

GitHub tag DOI License: MIT

AI Descartes: Combining Data and Theory for Derivable Scientific Discovery

This repository contains the code and the data used for the experiments in the paper Combining data and theory for derivable scientific discovery with AI-Descartes.

Visit our website for a general overview, references, and some introductory videos: → AI-Descartes website ←

system overview

Folders description:

  • data: contains the 3 datasets used in the paper (Kepler’s third law of planetary motion, Einstein’s time-dilation formula, Langmuir’s adsorption equation), the data points for 81 FSRD problems and the corresponding background theories (see data/README.md).
  • reasoning: contains the code for the Reasoning module of AI-Descartes (see reasoning/README.md)
  • symbolic-regression: contains the code for the Symbolic Regression module of AI-Descartes (see symbolic-regression/README.md)

How to cite

@article{AI_Descartes,
	title = {Combining data and theory for derivable scientific discovery with {AI}-{Descartes}},
	volume = {14},
	issn = {2041-1723},
	url = {https://doi.org/10.1038/s41467-023-37236-y},
	doi = {10.1038/s41467-023-37236-y},
	number = {1},
	journal = {Nature Communications},
	author = {Cornelio, Cristina and Dash, Sanjeeb and Austel, Vernon and Josephson, Tyler R. and Goncalves, Joao and Clarkson, Kenneth L. and Megiddo, Nimrod and El Khadir, Bachir and Horesh, Lior},
	month = apr,
	year = {2023},
	pages = {1777},
}