Self-Supervised Contrastive Learning Performs Non-Linear System Identification
Abstract
Self-supervised learning (SSL) approaches have brought tremendous success across many tasks and domains. It has been argued that these successes can be attributed to a link between SSL and identifiable representation learning: Temporal structure and auxiliary variables ensure that latent representations are related to the true underlying generative factors of the data. Here, we deepen this connection and show that SSL can perform system identification in latent space. We propose dynamics contrastive learning, a framework to uncover linear, switching linear and non-linear dynamics under a non-linear observation model, give theoretical guarantees and validate them empirically.
Cite
Text
Laiz et al. "Self-Supervised Contrastive Learning Performs Non-Linear System Identification." International Conference on Learning Representations, 2025.Markdown
[Laiz et al. "Self-Supervised Contrastive Learning Performs Non-Linear System Identification." International Conference on Learning Representations, 2025.](https://mlanthology.org/iclr/2025/laiz2025iclr-selfsupervised/)BibTeX
@inproceedings{laiz2025iclr-selfsupervised,
title = {{Self-Supervised Contrastive Learning Performs Non-Linear System Identification}},
author = {Laiz, Rodrigo González and Schmidt, Tobias and Schneider, Steffen},
booktitle = {International Conference on Learning Representations},
year = {2025},
url = {https://mlanthology.org/iclr/2025/laiz2025iclr-selfsupervised/}
}