https://github.com/erichson/koopmanae

Consistent Koopman Autoencoders
https://github.com/erichson/koopmanae

autoencoder forecasting machine-learning-algorithms time-series

Last synced: about 2 months ago
JSON representation

Consistent Koopman Autoencoders

• Host: GitHub
• URL: https://github.com/erichson/koopmanae
• Owner: erichson
• License: gpl-3.0
• Created: 2019-11-21T00:16:06.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2023-05-23T21:29:23.000Z (almost 3 years ago)
• Last Synced: 2024-07-11T08:58:42.125Z (almost 2 years ago)
• Topics: autoencoder, forecasting, machine-learning-algorithms, time-series
• Language: Python
• Homepage:
• Size: 13.9 MB
• Stars: 62
• Watchers: 4
• Forks: 21
• Open Issues: 2
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Consistent Koopman Autoencoders

Research code that demonstrates consistent Koopman autoencoders on the nonlinear pendulum with no friction.

Here is an example:

``` 

python driver.py --dataset pendulum --folder results_back_pendulum --bottleneck 6 --backward 1

```

You can also create a baseline model:

```

python driver.py --dataset pendulum --folder results_pendulum --bottleneck 6 --backward 0

```

Use the following code to plot the results:

```

python plot_pred_error.py

```



##  Reference

[Forecasting Sequential Data Using Consistent Koopman Autoencoders](https://arxiv.org/pdf/2003.02236.pdf)