Empirical Inference Article 2003

Dealing with large Diagonals in Kernel Matrices

In kernel methods, all the information about the training data is contained in the Gram matrix. If this matrix has large diagonal values, which arises for many types of kernels, then kernel methods do not perform well: We propose and test several methods for dealing with this problem by reducing the dynamic range of the matrix while preserving the positive definiteness of the Hessian of the quadratic programming problem that one has to solve when training a Support Vector Machine, which is a common kernel approach for pattern recognition.

Author(s): Weston, J. and Schölkopf, B. and Eskin, E. and Leslie, C. and Noble, WS.
Journal: Annals of the Institute of Statistical Mathematics
Volume: 55
Number (issue): 2
Pages: 391-408
Year: 2003
Month: June
Day: 0
Bibtex Type: Article (article)
DOI: 10.1007/BF02530507
Digital: 0
Electronic Archiving: grant_archive
Language: en
Organization: Max-Planck-Gesellschaft
School: Biologische Kybernetik
Links:

BibTex

@article{1866,
  title = {Dealing with large Diagonals in Kernel Matrices},
  journal = {Annals of the Institute of Statistical Mathematics},
  abstract = {In kernel methods, all the information about the training data is contained in the Gram matrix. If this matrix has large diagonal values, which arises for many types of kernels, then kernel methods do not perform well: We propose and test several methods for dealing with this problem by reducing the dynamic range of the matrix while preserving the positive definiteness of the Hessian of the quadratic programming problem that one has to solve when training a Support Vector Machine, which is a common kernel approach for pattern recognition.},
  volume = {55},
  number = {2},
  pages = {391-408},
  organization = {Max-Planck-Gesellschaft},
  school = {Biologische Kybernetik},
  month = jun,
  year = {2003},
  slug = {1866},
  author = {Weston, J. and Sch{\"o}lkopf, B. and Eskin, E. and Leslie, C. and Noble, WS.},
  month_numeric = {6}
}