kmatrix.m

this a SVM toolbox,it is very useful for someone who just learn SVM.In order to be undestood easily,

% KMATRIX computes kernel matrix for given data.
% [K] = kmatrix( data, ker, arg )
% [K] = kmatrix( dataA, dataB, ker, arg )
%
% Two cases:
%  1) K(i,j) = kernel( data(:,i), data(:,j)) i,j=1,...,N  
% or 
%  2) K(i,j) = kernel( dataA(:,i), dataB(:,j)) i=1,...,N1, j=1,...,N2
%
% Example: for 'linear' kernel it returns 1) data'*data 
% or 2) dataA'*dataB.
%
% Inputs:
%  1) data [D x N ] matrix of N training D-dimensional patterns.
% or 
%  2) dataA [D x N1 ] matrix of N1 training D-dimensional patterns.
%     dataB [D x N2 ] matrix of N2 training D-dimensional patterns.
%
%   ker [string] kernel identifier.
%   arg [real] kernel argument.
%
%     ker      arg      Kernel function
%    -----------------------------------------------------------------
%     'linear' []       Linear kernel: k(a,b) = a'*b
%     'poly'   d [int]  Polynom: k(a,b)=(a'*b +1)^d
%     'rbf'    s [real] Radial Basis Functions: k(a,b)=exp^(0.5*||a-b||^2/s^2)
%
%  Outputs:
%   1) K [N x N] kernel matrix.
%  or
%   2) K [N1 x N2] kernel matrix.
%

% Statistical Pattern Recognition Toolbox, Vojtech Franc, Vaclav Hlavac
% (c) Czech Technical University Prague, http://cmp.felk.cvut.cz
% Written Vojtech Franc (diploma thesis) 02.11.1999
%
% Modifications.
%  13-sep-2002, VF
%  21-October-2001, V.Franc
%  2-October-2001, V.Franc, created.