Genetic Equation Discovery

My goal was to explore machine learning equation discovery in the context of physics simulations. This project uses genetic programming to discover mathematical equations from data, with a focus on symbolic regression using the gplearn library. Example datasets and scripts are provided for experimentation and demonstration.

Here’s a GIF showcasing one of the results I obtained:

On the left, you can see a physically accurate pendulum simulated with its true physics equation. On the right, you see the formula automatically generated by gplearn.

Project Structure

├── datasets/                # Example datasets (CSV)
├── exemples/                # Example scripts (e.g., pendulum.py)
├── output/                  # Output files (plots, gifs, etc.)
├── src/                     # Source code (dataset generation, discovery)
├── README.md                # Project documentation

Installation

1. Clone the repository:

   git clone https://github.com/AngelLagr/genetic-eq-discovery.git
   cd genetic-eq-discovery

2. Install dependencies:

   pip install numpy pandas gplearn scikit-learn

Usage

To run the pendulum example and discover equations using genetic programming:

python -m examples.pendulum

This command runs the script as a module, ensuring imports work correctly.

License

Apache 2.0