Feedforward neural nets as discretization schemes for ODEs and DAEs
feedforward neural netsindex-3iterative algorithmslearning strategymultibody system dynamicsneural netsnonlinear least squares problemparallel computationparallel shooting-type algorithm
Implicit ordinary differential equations, differential-algebraic equations (34A09) Nonlinear ordinary differential equations and systems (34A34) Numerical methods for initial value problems involving ordinary differential equations (65L05) Parallel numerical computation (65Y05) Learning and adaptive systems in artificial intelligence (68T05)
Because of their neurophysical origin neural nets can be studied for classification tasks, approximation properties or iterative algorithms. They can be interpreted as a distributed or massively parallel computer, where each unit accumulates a scalar product and computes a one-dimensional nonlinear activation function. A supervised learning strategy defines a nonlinear least squares problem, which is solved by gradient techniques such as backpropagation. Interpreting the weights in a net as state variable feedforward neural nets can be designed as numerical discretization schemes for ordinary differential equations (ODEs) and differential-algebraic equations (DAEs). The net architecture for the implicit Euler scheme is presented here and some experiments are given. The net approach is of interest for the overdetermined index-3 approach of DAEs from multibody system dynamics. More generally, these nets determine a parallel shooting-type algorithm.
- The numerical solution of linear ordinary differential equations by feedforward neural networks
- Solution of nonlinear ordinary differential equations by feedforward neural networks
- scientific article; zbMATH DE number 1221486
- A numerical method for solving ordinary differential equations by using neural network
- A nonlinear truck model and its treatment as a multibody system
- An approach to nonlinear programming
- Convergence aspects of step-parallel iteration of Runge-Kutta methods
- Differential-algebraic equations in vehicle system dynamics
- scientific article; zbMATH DE number 50395 (Why is no real title available?)
- scientific article; zbMATH DE number 1821102 (Why is no real title available?)
- Ill-Conditioning in Neural Network Training Problems
- Parallel-iterated Runge-Kutta methods for stiff ordinary differential equations
- The numerical solution of linear ordinary differential equations by feedforward neural networks
- Solution of nonlinear ordinary differential equations by feedforward neural networks
- Learning to differentiate
- Regression-based neural network training for the solution of ordinary differential equations
- AI 2005: Advances in Artificial Intelligence
- Numerical Analysis for Convergence of a Sample-Wise Backpropagation Method for Training Stochastic Neural Networks
- Physics-informed neural networks and functional interpolation for stiff chemical kinetics
- Implementation and (inverse modified) error analysis for implicitly templated ODE-nets
- Stochastic operator network: a stochastic maximum principle based approach to operator learning
This page was built for publication: Feedforward neural nets as discretization schemes for ODEs and DAEs
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1372058)
