Premium
A Toy Model to Investigate Stability of AI‐Based Dynamical Systems
Author(s) -
Balogh B.,
SaintMartin D.,
Ribes A.
Publication year - 2021
Publication title -
geophysical research letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.007
H-Index - 273
eISSN - 1944-8007
pISSN - 0094-8276
DOI - 10.1029/2020gl092133
Subject(s) - artificial neural network , computer science , dimension (graph theory) , latin hypercube sampling , dynamical systems theory , feedforward neural network , instability , feed forward , lorenz system , stability (learning theory) , replicate , recurrent neural network , sample (material) , orbit (dynamics) , artificial intelligence , mathematics , machine learning , physics , chaotic , control engineering , statistics , quantum mechanics , aerospace engineering , mechanics , pure mathematics , monte carlo method , thermodynamics , engineering
The development of atmospheric parameterizations based on neural networks is often hampered by numerical instability issues. Previous attempts to replicate these issues in a toy model have proven ineffective. We introduce a new toy model for atmospheric dynamics, which consists in an extension of the Lorenz’63 model to a higher dimension. While feedforward neural networks trained on a single orbit can easily reproduce the dynamics of the Lorenz’63 model, they fail to reproduce the dynamics of the new toy model, leading to unstable trajectories. Instabilities become more frequent as the dimension of the new model increases, but are found to occur even in very low dimension. Training the feedforward neural network on a different learning sample, based on Latin Hypercube Sampling, solves the instability issue. Our results suggest that the design of the learning sample can significantly influence the stability of dynamical systems driven by neural networks.