Parsimonious neural networks

Design and train neural networks in conjunction with genetic algorithms to discover equations directly from data

Launch Tool

You must login before you can run this tool.

Version 1.2 - published on 22 Apr 2021

doi:10.21981/6K7M-GB17 cite this

Open source: license | download

View All Supporting Documents



Published on


This tool demonstrates the use of neural networks and genetic algorithms to discover scientific equations. We do this by training models that not only reproduce training and testing data accurately, but also achieve the simplest, most interpretable model possible. In this tool we will observe data of a particle moving under a non-linear external potential and aim to learn the underlying equations directly from the data.

Powered by

LAMMPS, Keras and DEAP

Cite this work

Researchers should cite this work as follows:

  • Saaketh Desai, Alejandro Strachan (2021), "Parsimonious neural networks," (DOI: 10.21981/6K7M-GB17).

    BibTex | EndNote