Stability of implicit neural networks for long-term forecasting in dynamical systems

05/26/2023
by   Leon Migus, et al.
0

Forecasting physical signals in long time range is among the most challenging tasks in Partial Differential Equations (PDEs) research. To circumvent limitations of traditional solvers, many different Deep Learning methods have been proposed. They are all based on auto-regressive methods and exhibit stability issues. Drawing inspiration from the stability property of implicit numerical schemes, we introduce a stable auto-regressive implicit neural network. We develop a theory based on the stability definition of schemes to ensure the stability in forecasting of this network. It leads us to introduce hard constraints on its weights and propagate the dynamics in the latent space. Our experimental results validate our stability property, and show improved results at long-term forecasting for two transports PDEs.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
06/15/2022

Learning to Accelerate Partial Differential Equations via Latent Global Evolution

Simulating the time evolution of Partial Differential Equations (PDEs) o...
research
06/22/2020

Phase space learning with neural networks

This work proposes an autoencoder neural network as a non-linear general...
research
09/03/2021

Semi-Implicit Neural Solver for Time-dependent Partial Differential Equations

Fast and accurate solutions of time-dependent partial differential equat...
research
09/04/2023

Dynamical Stability of Threshold Networks over Undirected Signed Graphs

In this paper we study the dynamic behavior of threshold networks on und...
research
10/08/2019

Implicit Neural Solver for Time-dependent Linear PDEs with Convergence Guarantee

Fast and accurate solution of time-dependent partial differential equati...
research
07/15/2022

Least squares solvers for ill-posed PDEs that are conditionally stable

This paper is concerned with the design and analysis of least squares so...
research
09/18/2022

Deep Convolutional Architectures for Extrapolative Forecast in Time-dependent Flow Problems

Physical systems whose dynamics are governed by partial differential equ...

Please sign up or login with your details

Forgot password? Click here to reset