Two Newton methods on the manifold of fixed-rank matrices endowed with Riemannian quotient geometries.
Pierre-Antoine AbsilLuca AmodeiGilles MeyerPublished in: Comput. Stat. (2014)
Keyphrases
- data points
- euclidean space
- symmetric positive definite
- riemannian manifolds
- geodesic distance
- fisher information
- riemannian metric
- euclidean distance
- manifold learning
- tangent space
- lie group
- singular values
- finite dimensional
- positive definite
- machine intelligence
- covariance matrices
- shape space
- distribution function
- shape analysis
- singular value decomposition
- diffusion tensor
- log euclidean
- essential matrices
- kernel matrix
- information geometry
- vector space