This repository contains the official implementations of the models and experiments presented in the pre-print paper: Frion, Anthony, et al. "Augmented Invertible Koopman Autoencoder for long-term time series forecasting." arXiv preprint arXiv:2503.12930 (2025).

Basic usage: Open any of the three notebooks in a Google Colab session and follow the instructions. The usage of a GPU is highly recommended for the computations to be realised in a reasonable time.

Alternatively, you may open the notebooks in any local framework (e.g. Jupyter notebooks), in which case you might need to adjust the paths to the different files in order for everything to work properly.