Saved in:
Bibliographic Details
Main Authors: Hjikakou, Kyriakos, Cartagena, Juan Diego Cardenas, Sabatelli, Matthia
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.18954
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914006339420160
author Hjikakou, Kyriakos
Cartagena, Juan Diego Cardenas
Sabatelli, Matthia
author_facet Hjikakou, Kyriakos
Cartagena, Juan Diego Cardenas
Sabatelli, Matthia
contents This paper investigates the generalisability of Koopman-based representations for chaotic dynamical systems, focusing on their transferability across prediction and control tasks. Using the Lorenz system as a testbed, we propose a three-stage methodology: learning Koopman embeddings through autoencoding, pre-training a transformer on next-state prediction, and fine-tuning for safety-critical control. Our results show that Koopman embeddings outperform both standard and physics-informed PCA baselines, achieving accurate and data-efficient performance. Notably, fixing the pre-trained transformer weights during fine-tuning leads to no performance degradation, indicating that the learned representations capture reusable dynamical structure rather than task-specific patterns. These findings support the use of Koopman embeddings as a foundation for multi-task learning in physics-informed machine learning. A project page is available at https://kikisprdx.github.io/.
format Preprint
id arxiv_https___arxiv_org_abs_2508_18954
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle On the Generalisation of Koopman Representations for Chaotic System Control
Hjikakou, Kyriakos
Cartagena, Juan Diego Cardenas
Sabatelli, Matthia
Machine Learning
This paper investigates the generalisability of Koopman-based representations for chaotic dynamical systems, focusing on their transferability across prediction and control tasks. Using the Lorenz system as a testbed, we propose a three-stage methodology: learning Koopman embeddings through autoencoding, pre-training a transformer on next-state prediction, and fine-tuning for safety-critical control. Our results show that Koopman embeddings outperform both standard and physics-informed PCA baselines, achieving accurate and data-efficient performance. Notably, fixing the pre-trained transformer weights during fine-tuning leads to no performance degradation, indicating that the learned representations capture reusable dynamical structure rather than task-specific patterns. These findings support the use of Koopman embeddings as a foundation for multi-task learning in physics-informed machine learning. A project page is available at https://kikisprdx.github.io/.
title On the Generalisation of Koopman Representations for Chaotic System Control
topic Machine Learning
url https://arxiv.org/abs/2508.18954