Title :
Multi resolution least squares SVM solver
Author :
Schouten, T. ; Suykens, J.A.K. ; De Moor, B.
Author_Institution :
Dept. Electrotechniek, Katholieke Univ. Leuven, Belgium
Abstract :
Discusses the use of multi resolution analysis (MRA) for fast approximate solution of large linear sets of equations arising in least squares support vector machine (LS-SVM) problems. When LS-SVMs are used on a low dimensional input space, the matrix of the linear set exhibits a structure that leads to a sparse approximation in the wavelet domain. The amount of structure decreases with increasing dimensionality. We illustrate this principle by means of a small example
Keywords :
learning automata; least squares approximations; neural nets; quadratic programming; sparse matrices; wavelet transforms; least squares SVM solver; linear equation sets; low dimensional input space; multi resolution analysis; nonlinear function estimation; quadratic programming; sparse approximation; support vector machine problems; wavelet domain; Ear; Equations; Least squares approximation; Least squares methods; Matrix decomposition; Multiresolution analysis; Sparse matrices; Support vector machines; Wavelet domain; Wavelet transforms;
Conference_Titel :
Circuits and Systems, 2000. Proceedings of the 43rd IEEE Midwest Symposium on
Conference_Location :
Lansing, MI
Print_ISBN :
0-7803-6475-9
DOI :
10.1109/MWSCAS.2000.951425
