mggpy 0.0.3

Multi-Gene Genetic Programming for system identification and regression

MGGP (Multigene Genetic Programming)

A flexible and extensible implementation of Multigene Genetic Programming (MGGP) focused on system identification and symbolic modeling, supporting:

• Regression (static and dynamic systems)
• Classification tasks
• Dynamic models in NARX and FIR formulations
• SISO, MISO, and MIMO system configurations

---

📦 Version: v0.0.3 (Latest Release)

This version introduces important improvements in modeling capabilities, code clarity, and project structure, moving the project closer to a fully packaged Python library.

🚀 What's new?

• ✅ Improved readability of Free-Run and Multi-Shooting (MShooting) implementations

  • Clearer iteration flow
  • Explicit batch handling in MShooting
  • Consistent behavior across SISO, MISO, and MIMO

• ✅ Support for current input term u[k]
  Previously, models only considered delayed inputs (u[k-1], ..., u[k-lagMax]).
  Including u[k] significantly enhances the expressive power of the identified models.

• ✅ Advancements toward packaging as a Python library
  Internal refactoring and structural improvements preparing for future PyPI release.

• ✅ Improved symbolic simplification for SISO models
  Cleaner and more interpretable final equations.

• ✅ General code cleanup and notebook reorganization

---

🧠 Core Features

• Symbolic regression via Multigene Genetic Programming
• Support for dynamic system identification (NARX / FIR)
• Native handling of:
  • SISO (Single Input Single Output)
  • MISO (Multiple Input Single Output)
  • MIMO (Multiple Input Multiple Output)
• Free-run simulation and multi-shooting training strategies
• Interpretable model structures (explicit equations)

---

🚀 Quick Start

Before start to run any notebook, run the code below to install the mggp as a library:

pip install -e .

otherwise, you'll need use:

src.mggp import MGGP

Check out the example notebook to get started:

📓 01_MGGP_Regression.ipynb - Demonstrates how to configure and use MGGP for Regression problems, and

📓 01_MGGP_Classifier.ipynb - Demonstrates how to configure and use MGGP for Classification problems.

📋 Requirements

The requirements are descript in the requirements.txt

This implementation is based on the MGGP methodology described in the following papers:

Foundational Works

Multi-Gene Genetic Programming for Nonlinear MIMO Modeling of F16 Aircraft Ground Vibrations

DOS SANTOS, Rafael Ávila et al. Multi-Gene Genetic Programming para Modelagem MIMO Nao Linear das Vibraçoes de uma Aeronave F16 no Solo.*

📄 Download PDF

A Novel MIMO Multi-Gene Genetic Programming Approach for Interpretable NARX Models: an application to vehicle state estimation

HENRIQUE GROENNER BARBOSA, Bruno et al. A Novel MIMO Multi-Gene Genetic Programming Approach for Interpretable NARX Models: an application to vehicle state estimation. Henrique and Ávila Santos, Rafael and Correa Victorino, Alessandro and Askari, Hassan and Xu, Nan, A Novel MIMO Multi-Gene Genetic Programming Approach for Interpretable NARX Models: an application to vehicle state estimation.*

📄 Download PDF

🤝 Contributing

Contributions are welcome! Feel free to open issues or submit pull requests.