On some neural network architectures that can represent viscosity solutions of certain high dimensional Hamilton-Jacobi partial differential equations.
Jérôme DarbonTingwei MengPublished in: J. Comput. Phys. (2021)
Keyphrases
- partial differential equations
- hamilton jacobi
- finite difference
- high dimensional
- boundary value problem
- differential equations
- numerical solution
- numerical algorithms
- image denoising
- anisotropic diffusion
- level set
- image processing
- eikonal equation
- multiscale
- image enhancement
- mathematical morphology
- numerical methods
- nonlinear partial differential equations
- diffusion equation
- high order
- optimal solution
- sufficient conditions
- level set method
- medial axis
- image compression
- feature space
- object recognition