Title :
Sparse Representation in Kernel Machines
Author :
Hongwei Sun ; Qiang Wu
Author_Institution :
Sch. of Math. Sci., Univ. of Jinan, Jinan, China
Abstract :
We study the properties of least square kernel regression with ℓ1 coefficient regularization. The kernels can be flexibly chosen to be either positive definite or indefinite. Asymptotic learning rates are deduced under smoothness condition on the kernel. Sparse representation of the solution is characterized theoretically. Empirical simulations and real applications indicate that both good learning performance and sparse representation could be guaranteed.
Keywords :
learning (artificial intelligence); least squares approximations; regression analysis; ℓ1 coefficient regularization; asymptotic learning rates; kernel machines; least square kernel regression; smoothness condition; sparse representation; Approximation methods; Compressed sensing; Kernel; Learning systems; Mathematical model; Polynomials; Probability distribution; $ell _{1}$ regularization; ℓ₁ regularization; Indefinite kernel; kernel machine; learning theory; regression; sparsity; sparsity.;
Journal_Title :
Neural Networks and Learning Systems, IEEE Transactions on
DOI :
10.1109/TNNLS.2014.2375209
