Dynamical Systems–Based Neural Networks
DOI10.1137/22m1527337arXiv2210.02373MaRDI QIDQ6181900
Elena Celledoni, Carola-Bibiane Schönlieb, Davide Murari, Ferdia Sherry, Brynjulf Owren
Publication date: 20 December 2023
Published in: SIAM Journal on Scientific Computing (Search for Journal in Brave)
Full work available at URL: https://arxiv.org/abs/2210.02373
dynamical systemsneural networksuniversal approximation theoremLipschitz networksstructure-preserving deep learning
Artificial neural networks and deep learning (68T07) Numerical methods for initial value problems involving ordinary differential equations (65L05) Multistep, Runge-Kutta and extrapolation methods for ordinary differential equations (65L06) Discretization methods and integrators (symplectic, variational, geometric, etc.) for dynamical systems (37M15)
Cites Work
- Unnamed Item
- Unnamed Item
- Switching in systems and control
- A high-order conservative Patankar-type discretisation for stiff systems of production--destruction equations
- \(L^p\) approximation of maps by diffeomorphisms
- Non-local decomposition of vector fields
- Deep neural networks motivated by partial differential equations
- Sympnets: intrinsic structure-preserving symplectic networks for identifying Hamiltonian systems
- Learning Hamiltonians of constrained mechanical systems
- Order theory for discrete gradient methods
- The modern mathematics of deep learning
- Convolutional proximal neural networks and plug-and-play algorithms
- Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
- A proposal on machine learning via dynamical systems
- Reducibility and contractivity of Runge-Kutta methods revisited
- Control on the manifolds of mappings with a view to the deep learning
- Deep limits of residual neural networks
- Locally-symplectic neural networks for learning volume-preserving dynamics
- Simulating Hamiltonian Dynamics
- Splitting methods
- Basic problems in stability and design of switched systems
- Structure-preserving deep learning
- Positivity-preserving methods for ordinary differential equations
- Deep Hamiltonian networks based on symplectic integrators
- Solving inverse problems using data-driven models
- Geometric Numerical Integration
- Hamiltonian Deep Neural Networks Guaranteeing Nonvanishing Gradients by Design
- Neural ODE Control for Classification, Approximation, and Transport
- PDE-based group equivariant convolutional neural networks
- Deep learning via dynamical systems: an approximation perspective
