Riemannian Trust Regions with Finite-Difference Hessian Approximations are Globally Convergent.
Nicolas BoumalPublished in: GSI (2015)
Keyphrases
- finite difference
- globally convergent
- numerical analysis
- numerical solution
- finite element
- partial differential equations
- autocalibration
- variational inequalities
- global convergence
- newton method
- level set
- shape analysis
- augmented lagrangian
- diffusion equation
- parameter space
- input image
- euclidean space
- image segmentation
- single image
- image features
- image analysis
- pairwise