Material covered

  • Kernels in local regression -- Nadaraya-Watson
  • SVM dual, where to plug in the kernel
  • Gram matrix and why it should be positive semi definite
  • Eigen decomposition, embedding in feature space
  • Polynomial kernels from first principles
  • Rules for creating valid kernels by combining known kernels

Last modified: Thursday, 27 September 2007, 12:07 PM