Regularization networks with indefinite kernels (Q1935751)
From MaRDI portal
scientific article
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | Regularization networks with indefinite kernels |
scientific article |
Statements
Regularization networks with indefinite kernels (English)
0 references
19 February 2013
0 references
The author considers the least squares regression problem by coefficient-based regularization schemes with \(\ell_2\) norm square as the penalty generated by indefinite kernels. The key interesting idea is to introduce a pair of reproducing kernel Hilbert spaces for a singular value decomposition of the kernel integral. By methods from functional analysis, the author derives capacity independent error bounds and learning rates.
0 references
learning theory
0 references
coefficient-based regularization
0 references
indefinite kernel
0 references
least square regression
0 references
reproducing kernel Hilbert space
0 references
capacity independent error bound
0 references
learning rates
0 references
