Handy package for training output constrained neural networks. As of now, this package only contains constrained extreme learning machines (CELM), constrained multi layer neural networks will be added in the future.
The source code is contained in the src folder. The implementation of the (C)ELM can be found in src/learner.py.
To install the package:
pip install -r requirements.txt
To run the synthetic examples:
python examples\flat_gaussian.py
python examples\monotone_quadratic.py
python examples\ode.py
To run a more complex example that involves learning a Lyapunov function, which in turn defines constraints for the learning process of a stable dynamical system:
python examples\stable_dynamical_system.py
For reference, take a look at this paper.
MIT license - See LICENSE.