Koopman Autoencoders Learn Representation Dynamics

Authors
Affiliation

Nishant Aswani

NYU Abu Dhabi, NYU Tandon

Saif Eddin Jabari

NYU Abu Dhabi, NYU Tandon

Published

September 23, 2025

The visuals below complement our paper and poster.

Figure 1: Koopman autoencoder (\(k=100\)) representation predictions for the Yin-Yang dataset.
Figure 2: Koopman autoencoder (\(k=10\)) representation predictions for the MNIST dataset.

Citation

For attribution, please cite this work as:
Aswani, Nishant, and Saif Eddin Jabari. 2025. “Koopman Autoencoders Learn Representation Dynamics.” arXiv Preprint arXiv:2505.12809, September. https://arxiv.org/abs/2505.12809.