lpdist.m

SVDD的工具箱

function D = lpdist(a,b,p,w)
%LPDIST Fast L_p distance 
%
%     D = LPDIST(A,B,P,W)
%
% Compute the L_p^p distance between data A and B in a fast(er) way than
% using myproxm.m using p=P. The features can also be weighted by
% weights W.
%
%    d(a,b) =  [sum_i abs(a_i-b_i)^p ]
%
% See also: lpball_dd

% Copyright: D.M.J. Tax, [e-mail]
% Faculty EWI, Delft University of Technology
% P.O. Box 5031, 2600 GA Delft, The Netherlands
if nargin<4
	w = [];
end
if nargin<3
	p = 2;
end

% size of the data:
[m,d] = size(a);
% size of the means:
n = size(b,1);
D = zeros(m,n);

if isempty(w)
	for i=1:n
		D(:,i) = sum( abs(a-repmat(b(i,:),m,1)).^p ,2);
	end
else
	for i=1:n
		D(:,i) = sum( repmat(w(i,:),m,1).*abs(a-repmat(b(i,:),m,1)).^p ,2);
	end
end



return