Blended kernel approximation in the ℋ-matrix techniques.
Wolfgang HackbuschBoris N. KhoromskijPublished in: Numer. Linear Algebra Appl. (2002)
Keyphrases
- positive definite
- singular vectors
- singular value decomposition
- gram matrix
- condition number
- positive semidefinite
- kernel function
- polynomial kernels
- series expansion
- error bounds
- matrix representation
- positive semi definite
- dot product
- approximation algorithms
- approximation error
- similarity measure
- kernel methods
- online learning
- data matrix
- sparse matrix
- kernel pca
- kernel learning
- input space
- kernel matrices
- closed form
- higher education
- pairwise