Skip to main content

Online and Window Dynamic Mode Decomposition algorithms

Project description

odmd

Python/Matlab implementation of online dynamic mode decomposition (Online DMD) and window dynamic mode decomposition (Window DMD) algorithms proposed in this paper.

Hightlights

  • The online algorithm is optimal in terms of both time and space complexity. Its time complxity is O(n^2), where n is the state dimension. This is much faster than standard algorithm O(n^3). Its space complexity is O(n^2), which is much more efficient than standard algorithm O(n*T), where T is the number of measurements (T >> n, and will go to infinity in online applications)
  • It finds the exact optimal solution, without any approximation (Unlike stochastic gradient descent).
  • It is a very general algorithm, and applies to problems other than fluid dynamics, dynamical systems, and control. If we have a function (or a dynamical model) y = M*phi(x), where phi is some known arbitrary nonlinear vector-valued function, x, y are vectors, M is an unknown matrix, and measurement x, y comes in real-time, then this algorithm updates the model M optimally. This form is very general, and can represent a very large class of models such as LTI, Lorenz attractor, polynomial system, vector AR, and many more.
  • It can be used for model predictive control.
  • It has been successfully applied to flow control problems, and achived great real-time closed loop control. See this paper for details.

Algorithm Description

Here is a brief description of the proposed algorithms. For more details, please refer to the original paper.

Online DMD algorithm description

The algorithm is implemented in class OnlineDMD.

At time step k, define two matrix X(k) = [x(1),x(2),...,x(k)], Y(k) = [y(1),y(2),...,y(k)], that contain all the past snapshot pairs, where x(k), y(k) are the n dimensional state vector, y(k) = f(x(k)) is the image of x(k), f() is the dynamics. Here, if the (discrete-time) dynamics are given by z(k) = f(z(k-1)), then x(k), y(k) should be measurements corresponding to consecutive states z(k-1) and z(k).

An exponential weighting factor rho=sigma^2 (0<rho<=1) that places more weight on recent data can be incorporated into the definition of X(k) and Y(k) such that X(k) = [sigma^(k-1)*x(1),sigma^(k-2)*x(2),…,sigma^(1)*x(k-1),x(k)], Y(k) = [sigma^(k-1)*y(1),sigma^(k-2)*y(2),...,sigma^(1)*y(k-1),y(k)].

At time k+1, the matrices become X(k+1) = [x(1),x(2),…,x(k),x(k+1)], Y(k+1) = [y(1),y(2),…,y(k),y(k+1)]. We need to remember a new snapshot pair x(k+1), y(k+1). We can update the DMD matrix Ak = Yk*pinv(Xk) recursively by efficient rank-1 updating online DMD algorithm.

The time complexity (multiply–add operation for one iteration) is O(n^2), and space complexity is O(n^2), where n is the state dimension.

Window DMD algorithm description

The algorithm is implemented in class WindowDMD.

At time step k, define two matrix X(k) = [x(k-w+1),x(k-w+2),...,x(k)], Y(k) = [y(k-w+1),y(k-w+2),...,y(k)] that contain the recent w snapshot pairs from a finite time window, where x(k), y(k) are the n dimensional state vector, y(k) = f(x(k)) is the image of x(k), f() is the dynamics. Here, if the (discrete-time) dynamics are given by z(k) = f(z(k-1)), then x(k), y(k) should be measurements corresponding to consecutive states z(k-1) and z(k).

An exponential weighting factor rho=sigma^2 (0<rho<=1) that places more weight on recent data can be incorporated into the definition of X(k) and Y(k) such that X(k) = [sigma^(w-1)*x(k-w+1),sigma^(w-2)*x(k-w+2),…,sigma^(1)*x(k-1),x(k)], Y(k) = [sigma^(w-1)*y(k-w+1),sigma^(w-2)*y(k-w+2),…,sigma^(1)*y(k-1),y(k)].

At time k+1, the data matrices become X(k+1) = [x(k-w+2),x(k-w+3),…,x(k+1)], Y(k+1) = [y(k-w+2),y(k-w+3),…,y(k+1)]. We need to forget the oldest snapshot pair x(k-w+1),y(k-w+1), and remember the newest snapshot pair x(k+1),y(k+1). We can update the DMD matrix Ak = Yk*pinv(Xk) recursively by efficient rank-2 updating window DMD algroithm.

The time complexity (multiply–add operation for one iteration) is O(n^2), and space complexity is O(wn+2n^2), where n is the state dimension, and w is the window size.

Installation

Use pip

python3 -m pip install odmd

Manual install

Create virtual env if needed

python3 -m venv .venv
source .venv/bin/activate

Clone from github and install

git clone https://github.com/haozhg/odmd.git
cd odmd/
python3 -m pip install -e .

Test

To run tests

cd tests/
pip install r requirements.txt
python -m pytest .

Demos

See demo for python notebooks.

Authors

Hao Zhang
Clarence W. Rowley

Reference

If you you used these algorithms or this python package in your work, please consider citing

Zhang, Hao, Clarence W. Rowley, Eric A. Deem, and Louis N. Cattafesta. "Online dynamic mode decomposition for time-varying systems." SIAM Journal on Applied Dynamical Systems 18, no. 3 (2019): 1586-1609.

BibTeX

@article{zhang2019online,
  title={Online dynamic mode decomposition for time-varying systems},
  author={Zhang, Hao and Rowley, Clarence W and Deem, Eric A and Cattafesta, Louis N},
  journal={SIAM Journal on Applied Dynamical Systems},
  volume={18},
  number={3},
  pages={1586--1609},
  year={2019},
  publisher={SIAM}
}

Date created

April 2017

License

If you want to use this package, but find license permission an issue, pls contact me at haozhang at alumni dot princeton dot edu.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

odmd-0.1.0.tar.gz (6.7 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page