NeuralNetwork

NeuralNetwork Modelica library.

Library description

The NeuralNetwork library is developed in Modelica 4.0 and offers dense feed-forward artificial neural network structures in pure Modelica.

Example

Modelica model NeuralNetwork.Examples.HelloWorld replace polynomial Modelica equation

$$ f: [-1, 1] \to \mathbb{R}, f(u) = u^2 + 0.5 u - 2 $$

with a dense feed forward neural network with one input and one output layer.

The training of the neural network is done with TensorFlow. Training and export to Modelica is showcased in Jupyter notebook Example/helloWorld.ipynb.

Current release

Version 2.1.0

License

This package is licensed under the 3-Clause BSD License.

Development and contribution

The development is organized by Hochschule Bielefeld – University of Applied Sciences and Arts, Faculty of Engineering and Mathematics.

Contributor need to sign a contributor license agreement.

This package is based on NeuralNetwork v1.0 (2006-04-28).

Build Python packages

PythonToModelica is a Python package to help exporting artificial neural networks from Python to Modelica and used in Example/helloWorld.ipynb. See README.md for build instructions.

Issues

You may report any issues by using the Issue Tracker.