PyKoopman¶
PyKoopman is a Python package for computing data-driven approximations to the Koopman operator.
Data-driven approximation of Koopman operator¶
Given a nonlinear dynamical system,
the Koopman operator governs the temporal evolution of the measurement function. Unfortunately, it is an infinite-dimensional linear operator. Most of the time, one has to project the Koopman operator onto a finite-dimensional subspace that is spanned by user-defined/data-adaptive functions. If the system state is also contained in such subspace, then effectively, the nonlinear dynamical system is (approximately) linearized in a global sense.
Structure of PyKoopman¶
PyKoopman package is centered around the Koopman
class and KoopmanContinuous
class. It consists of two key components
observables
: a set of observables functions, which spans the subspace for projection.regressor
: the optimization algorithm to find the bestfit
for the projection of Koopman operator.
After Koopman
/KoopmanContinuous
object has been created, it must be fit to data, similar to a scikit-learn
model.
We design PyKoopman
such that it is compatible to scikit-learn
objects and methods as much as possible.
Example¶
Installation¶
Installing with pip¶
If you are using Linux or macOS you can install PyKoopman with pip:
pip install pykoopman
Installing from source¶
First clone this repository:
git clone https://github.com/dynamicslab/pykoopman
Then, to install the package, run
pip install .
If you do not have pip you can instead use
python setup.py install
If you do not have root access, you should add the --user
option to the above lines.
Documentation¶
The documentation for PyKoopman is hosted on Read the Docs.
Community guidelines¶
Contributing code¶
We welcome contributions to PyKoopman. To contribute a new feature please submit a pull request. To get started we recommend installing the packages in requirements-dev.txt
via
pip install -r requirements-dev.txt
This will allow you to run unit tests and automatically format your code. To be accepted your code should conform to PEP8 and pass all unit tests. Code can be tested by invoking
pytest
We recommed using pre-commit
to format your code. Once you have staged changes to commit
git add path/to/changed/file.py
you can run the following to automatically reformat your staged code
pre-commit -a -v
Note that you will then need to re-stage any changes pre-commit
made to your code.
Reporting issues or bugs¶
If you find a bug in the code or want to request a new feature, please open an issue.
References¶
Williams, Matthew O., Ioannis G. Kevrekidis, and Clarence W. Rowley. A data–driven approximation of the koopman operator: Extending dynamic mode decomposition. Journal of Nonlinear Science 25, no. 6 (2015): 1307-1346. [DOI]
Williams, Matthew O., Clarence W. Rowley, and Ioannis G. Kevrekidis. A kernel-based approach to data-driven Koopman spectral analysis. arXiv preprint arXiv:1411.2260 (2014). [DOI]
Brunton, Steven L., et al. Chaos as an intermittently forced linear system. Nature communications 8.1 (2017): 1-9. [DOI]
Kaiser, Eurika, J. Nathan Kutz, and Steven L. Brunton. Data-driven discovery of Koopman eigenfunctions for control. Machine Learning: Science and Technology 2.3 (2021): 035023. [DOI]
Lusch, Bethany, J. Nathan Kutz, and Steven L. Brunton. Deep learning for universal linear embeddings of nonlinear dynamics. Nature communications 9.1 (2018): 4950. [DOI]
Otto, Samuel E., and Clarence W. Rowley. Linearly recurrent autoencoder networks for learning dynamics. SIAM Journal on Applied Dynamical Systems 18.1 (2019): 558-593. [DOI]
Pan, Shaowu, Nicholas Arnold-Medabalimi, and Karthik Duraisamy. Sparsity-promoting algorithms for the discovery of informative Koopman-invariant subspaces. Journal of Fluid Mechanics 917 (2021). [DOI]
- Learning how to create observables
- Learning how to compute time derivatives
- Dynamic mode decomposition on two mixed spatial signals
- Dynamic mode decomposition with control on a 2D linear system
- Dynamic mode decomposition with control (DMDc) for a 128D system
- Dynamic mode decomposition with control on a high-dimensional linear system
- Successful examples of using Dynamic mode decomposition on PDE system
- Unsuccessful examples of using Dynamic mode decomposition on PDE system
- Extended DMD for Van Der Pol System
- Learning Koopman eigenfunctions on Slow manifold
- Comparing DMD and KDMD for Slow manifold dynamics
- Extended DMD with control for chaotic duffing oscillator
- Extended DMD with control for Van der Pol oscillator
- Hankel Alternative View of Koopman Operator for Lorenz System
- Hankel DMD with control for Van der Pol Oscillator
- Neural Network DMD on Slow Manifold
- EDMD and NNDMD for a simple linear system
- Sparisfying a minimal Koopman invariant subspace from EDMD for a simple linear system