[Pattern recognition] SVM Kernel Function

The following are common core function representations:

Linear Kernel)

Polynomial kernel)

Radial Basis Function)

Also calledGaussian Kernel)Because it can be seen as a form of the following core function:

A radial basis function is a real-value function that depends only on the distance of a specific point. Any function that satisfies the characteristics is called the radial vector function. The standard Euclidean distance is generally used, although other distance functions can also be used. Therefore, the other two commonly used kernel functions, power exponent kernel and Laplace kernel, also belong to radial basis kernel functions. In addition, less commonly used radial basis cores include ANOVA, secondary rational, multivariate secondary, and inverse multivariate secondary.
Exponential Kernel)
  Laplacian Kernel)
  ANOVA kernel (ANOVA kernel)
  Rational Quadratic Kernel)
  Multiquadric Kernel)
  Inverse multiquadric Kernel)
Another simple and practical solution is: Sigmoid core (sigmoid kernel)
The preceding methods are commonly used. Most of them are directly set in SVM, SVM-light, And ranksvm. There are other uncommon ones, such as wavelet kernels and Bayesian kernels, which can be specified by the Code.



(Reprinted please indicate the author and Source: http://blog.csdn.net/xiaowei_cqu is not allowed for commercial use)