| Presentation | 2012-03-09 A Note on Multi-Kernel Adaptive Learning Based on RKHS Projection Ryuichiro ISHII, Masahiro YUKAWA, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | A multi-kernel adaptive learning based on RKHS projection (MKAL-RKHS) is investigated. It is first shown that the existing kernel adaptive learning algorithms can be classified into two categories from the vector space projection viewpoint: RKHS projection and parameter-pace projection approaches. A batch approach to determining convex combination coefficients to multiple kernels by Wrapper method based on optimization with an Ivanov regularization is then introduced, and a MKAL-RKHS algorithm employing the combination coefficients is presented. In kernel adaptive learning, an unknown nonlinear function is approximated by a linear combination of multiple Gaussian functions centered respectively at selected data points. In MKAL-RKHS, the shapes of functions are all the same at any selected data points. On the other hand, in the multi-kernel adaptive learning based on parameter-space projection (MKAL-PS) [Yukawa 2010], the shapes of functions are automatically adjusted independently from point to point so that the shape of function at each point fits the unknown function. The MKAL-RKHS and MKAL-PS algorithms are applied to online prediction of nonlinear time series data, and their performance is evaluated by simulations. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | kernel adaptive filter / multiple kernels / reproducing kernel Hilbert space (RKHS) / orthogonal projection |
| Paper # | CAS2011-163,SIP2011-183,CS2011-155 |
| Date of Issue |
| Conference Information | |
| Committee | CAS |
|---|---|
| Conference Date | 2012/3/1(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Circuits and Systems (CAS) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | A Note on Multi-Kernel Adaptive Learning Based on RKHS Projection |
| Sub Title (in English) | |
| Keyword(1) | kernel adaptive filter |
| Keyword(2) | multiple kernels |
| Keyword(3) | reproducing kernel Hilbert space (RKHS) |
| Keyword(4) | orthogonal projection |
| 1st Author's Name | Ryuichiro ISHII |
| 1st Author's Affiliation | Dept. Electrical and Electronic Engineering, Niigata University() |
| 2nd Author's Name | Masahiro YUKAWA |
| 2nd Author's Affiliation | Dept. Electrical and Electronic Engineering, Niigata University |
| Date | 2012-03-09 |
| Paper # | CAS2011-163,SIP2011-183,CS2011-155 |
| Volume (vol) | vol.111 |
| Number (no) | 465 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |