PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

A Rapidly Computable String Kernel Based on the Discrete Fourier Transform: Derivation and Experiments
Chris Watkins and Juho Rousu
(2004) Working Paper. Department of Computer Science, Royal Holloway, Egham, UK.


This paper derives a new(?) string kernel based on the discrete Fourier transform. This kernel is rapid to compute, and measures a slightly different kind of similarity from dynamic alignment or spectrum kernels. A preliminary experiment

Postscript - Requires a viewer, such as GhostView
EPrint Type:Monograph (Working Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Computational, Information-Theoretic Learning with Statistics
Theory & Algorithms
ID Code:789
Deposited By:Chris Watkins
Deposited On:30 December 2004