Abstract:Despite the great success of SVM, it is usually difficult for users to select suitable kernels for SVM classifiers. Kernel learning has been developed to jointly learn both a kernel and an SVM classifier [1]. Most existing kernel learning approaches, eg,[2, 3, 4], employ the margin based formulation, equivalent to: mink, w, b, ξi 1 2 w2+ C∑ i ξi, st yi〈 φ (xi; k), w〉+ b+ ξi≥ 1, ξi≥ 0,(1) where k is the learned kernel which implicitly defines a transformation φ (·; k) to a feature space by k (xc, xd)=〈 φ (xc; k), φ (xd; k)〉,(w, b) is an SVM classifier, and xi, yi and ξi are input instances, labels and hinge losses. To make the problem trackable, the learned kernel is usually restricted to a parametric form k (θ)(·,·), where θ=[θi] i is the kernel parameter. The most common used form is a linear combination of multiple basis kernels, as k (θ)(·,·)=∑ m j= 1 θjkj (·,·), θj≥ 0.(2)Let γ denote the margin of the SVM classifier with k. It is well known that γ− 2= w2. Thus the term w2 makes the margin based formulation (1) prefer the kernel that results in an SVM classifier with a larger margin. However, the margin itself can not well describe the goodness of a kernel. Any kernel, even one with a bad performance, can have arbitrarily large margin by enlarging the kernel’s scaling, and may be selected to be the final solution [5]. Therefore the margin based kernel learning methods suffer from scaling problems. In linear combination cases, a remedy is to enforce a norm constraint on kernel parameters. Unfortunately, it is difficult to select suitable types of norm constraints, and with norm constraints the scaling problem also causes another initialization problem: different initial scalings of basis …

A Selection Means on the Parameter of Radius Basis Function

Novel Linear Search for Support Vector Machine Parameter Selection

Parameter Selection For Gaussian Radial Basis Function In Support Vector Machine Classification

A Gaussian radial basis function based feature selection algorithm

Random radial basis function kernel-based support vector machine

An Analytical Approach to Fast Parameter Selection of Gaussian Rbf Kernel for Support Vector Machine

A Kernel Function Optimization And Selection Algorithm Based On Cost Function Maximization

Bayesian Approach for Radial Kernel Parameter Tuning

On Radial Basis Function Net and Kernel Regression: Approximation Ability, Convergence Rate and Receptive Field Size

Training Radial Basis Function Classifiers with Gaussian Kernel Clustering and Fuzzy Decision Technique

Adaptive selection strategy of shape parameters for LRBF for solving partial differential equations

Multiclass Kernel Polarization and Its Application to Parameter Selection of RBF Kernel with Multiple Widths

Kernel Parameter Selection for Support Vector Machine Classification

Parameter Selection of SVM with Gaussian kernel

On the numerical rank of radial basis function kernels in high dimension

Efficient Model Selection for Support Vector Machine with Gaussian Kernel Function

Comparative Studies of Support Vector Regression Between Reproducing Kernel and Gaussian Kernel

Determination of the Spread Parameter in the Gaussian Kernel for Classification and Regression

A New Method for Determining the Parameter of Gaussian Kernel

Learning Kernels via Margin-and-Radius Ratios

Nonlinear Ridge Regression Modeling Method Based on Radial Basis Function and Its Simulation Research