gendatw.m

The pattern recognition matlab toolbox

%GENDATW Sample dataset by given weigths
%
%   B = GENDATW(A,V,N)
%
% INPUT
%   A    Dataset
%   V    Vector with weigths for each object in A
%   N    Number of objects to be generated (default size A);
%
% OUTPUT
%   B    Dataset
%
% DESCRIPTION
% The dataset A is sampled using the weigths in V as a prior distribution.

function b = gendatw(a,v,n)

isdataset(a);
if nargin < 3, n  = size(a,1); end

v = v./sum(v);
if any(v<0)
    error('Weights should be positive');
end

mins = 0;
nn = 0;
while(mins < 2)
	N = genclass(n,v);
	L = [];
	while any(N > 0)
	    L = [L find(N > 0)];
	    N = N-1;
	end
	b = a(L,:);
	mins = min(classsizes(b));
	nn = nn + 1;
	if nn > 100
		error('Problems with weighted subsampling. Please enlarge training set.')
	end
end

return