Repository containing the code to learn stable dynamical system encoding non-linearity within underlying manifold's curvature. Paper: https://arxiv.org/abs/2403.11948

Starting from a flat manifold, the manifold curvature adapts so the dynamical system streamlines match the demonstrated trajectories (red dots). On the left the first-order learned dynamical system. On the right the second-order learned dynamical system.

• Bernardo Fichera ([email protected])

In order to train a model

python(python3) -m examples.train_model --data <dataset_name>
(ipython) run examples/train_model.py --data <dataset_name>

where the line commands data can be used to set the training dataset. In oder to test a trained model

python(python3) -m examples.test_model --data <dataset_name>
(ipython) run examples/test_model.py --data <dataset_name>

where the line command data can be used load the model/dataset.

In order to install the package in .local run

pip(pip3) install .

For local installation in the current directory

pip(pip3) install -e .