mog_extend.m

支持向量域是近几年采用的一种较新的分类器

function w = mog_extend(w,x,n,maxiter)
%MOG_EXTEND Extend a MoG with one cluster
%
%     W = MOG_EXTEND(W,X,[],MAXITER)
%
% Extend a Mixture of Gaussians model W to data X with one extra cluster
% for the target class. Maximally MAXITER EM-updates are made.
%
%     W = MOG_UPDATE(W,X,N)
%
% If input parameter N is given, you can specify if it should be for the
% target, outlier or both classes:
%   N = [0 1] : only extra outlier cluster
%   N = [1 1] : both target and outlier cluster.
%
% See also: mog_dd, mog_init, mog_P, mog_update

% 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
	maxiter = 25;
end
if nargin<3 | isempty(n)
	n = [1 0]
end

% initialize
if length(n)<2
	n = [n 0];
end

% get data:
dat = w.data;
[xt,xo] = target_class(x);
Pt = zeros(size(xt,1),1);
Po = zeros(size(xt,1),1);

% update the target class clusters:
if (n(1)>0) & ~isempty(xt)
	[dat.mt,dat.ict,dat.pt] = mogEMextend(+xt,dat.covtype,dat.mt,dat.ict,dat.pt,maxiter);
end
% update the outlier class clusters:
if (n(2)>0) & ~isempty(xo)
	[dat.mo,dat.ico,dat.po] = mogEMextend(+xo,dat.covtype,dat.mo,dat.ico,dat.po,maxiter);
end

% store it again:
w.data = dat;

return