Improved deep learning of chaotic dynamical systems with multistep penalty losses

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chakraborty, Dibyajyoti, Chung, Seung Whan, Chattopadhyay, Ashesh, Maulik, Romit
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866913535983878144
author Chakraborty, Dibyajyoti
Chung, Seung Whan
Chattopadhyay, Ashesh
Maulik, Romit
author_facet Chakraborty, Dibyajyoti
Chung, Seung Whan
Chattopadhyay, Ashesh
Maulik, Romit
contents Predicting the long-term behavior of chaotic systems remains a formidable challenge due to their extreme sensitivity to initial conditions and the inherent limitations of traditional data-driven modeling approaches. This paper introduces a novel framework that addresses these challenges by leveraging the recently proposed multi-step penalty (MP) optimization technique. Our approach extends the applicability of MP optimization to a wide range of deep learning architectures, including Fourier Neural Operators and UNETs. By introducing penalized local discontinuities in the forecast trajectory, we effectively handle the non-convexity of loss landscapes commonly encountered in training neural networks for chaotic systems. We demonstrate the effectiveness of our method through its application to two challenging use-cases: the prediction of flow velocity evolution in two-dimensional turbulence and ocean dynamics using reanalysis data. Our results highlight the potential of this approach for accurate and stable long-term prediction of chaotic dynamics, paving the way for new advancements in data-driven modeling of complex natural phenomena.
format Preprint
id arxiv_https___arxiv_org_abs_2410_05572
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Improved deep learning of chaotic dynamical systems with multistep penalty losses
Chakraborty, Dibyajyoti
Chung, Seung Whan
Chattopadhyay, Ashesh
Maulik, Romit
Machine Learning
Artificial Intelligence
Dynamical Systems
Predicting the long-term behavior of chaotic systems remains a formidable challenge due to their extreme sensitivity to initial conditions and the inherent limitations of traditional data-driven modeling approaches. This paper introduces a novel framework that addresses these challenges by leveraging the recently proposed multi-step penalty (MP) optimization technique. Our approach extends the applicability of MP optimization to a wide range of deep learning architectures, including Fourier Neural Operators and UNETs. By introducing penalized local discontinuities in the forecast trajectory, we effectively handle the non-convexity of loss landscapes commonly encountered in training neural networks for chaotic systems. We demonstrate the effectiveness of our method through its application to two challenging use-cases: the prediction of flow velocity evolution in two-dimensional turbulence and ocean dynamics using reanalysis data. Our results highlight the potential of this approach for accurate and stable long-term prediction of chaotic dynamics, paving the way for new advancements in data-driven modeling of complex natural phenomena.
title Improved deep learning of chaotic dynamical systems with multistep penalty losses
topic Machine Learning
Artificial Intelligence
Dynamical Systems
url https://arxiv.org/abs/2410.05572