kpca_map.m

来自「KPCA是一种非线性的盲源分离方法」· M 代码 · 共 29 行

function y = kpca_map(x,basis)
%KPCA_MAP maps input vectors onto the kpca subspace.
%
%   usage
%      psi = kpca_map(x,basis);
%
%   input
%      x        matrix of column vectors, containing the input
%               vectors
%      basis    a kpca basis, generated by kpca_calc (see there)
%
%   output
%      y        matrix of column vectors, containing the mapped
%               input vectors
%
%   see also
%      kpca_plot, kpca_calc
%
%   STH * 12MAR2002

% column of xs
xsc = size(basis.xs,2);

% calc kernel matrix for xs and x
Kxsx = kpca_matrix(basis.xs,x,basis.kernel);

% note, that "repmat..." ensures centering of xs
y = basis.V'*(Kxsx - repmat(sum(Kxsx,1),[xsc 1])/xsc);