camir-aes2014: toolboxes/FullBNT-1.0.7/KPMtools/sample

first hg version after svn

rev	line source
wolffd@0	1 function M = sample_discrete(prob, r, c)
wolffd@0	2 % SAMPLE_DISCRETE Like the built in 'rand', except we draw from a non-uniform discrete distrib.
wolffd@0	3 % M = sample_discrete(prob, r, c)
wolffd@0	4 %
wolffd@0	5 % Example: sample_discrete([0.8 0.2], 1, 10) generates a row vector of 10 random integers from {1,2},
wolffd@0	6 % where the prob. of being 1 is 0.8 and the prob of being 2 is 0.2.
wolffd@0	7
wolffd@0	8 n = length(prob);
wolffd@0	9
wolffd@0	10 if nargin == 1
wolffd@0	11 r = 1; c = 1;
wolffd@0	12 elseif nargin == 2
wolffd@0	13 c == r;
wolffd@0	14 end
wolffd@0	15
wolffd@0	16 R = rand(r, c);
wolffd@0	17 M = ones(r, c);
wolffd@0	18 cumprob = cumsum(prob(:));
wolffd@0	19
wolffd@0	20 if n < r*c
wolffd@0	21 for i = 1:n-1
wolffd@0	22 M = M + (R > cumprob(i));
wolffd@0	23 end
wolffd@0	24 else
wolffd@0	25 % loop over the smaller index - can be much faster if length(prob) >> r*c
wolffd@0	26 cumprob2 = cumprob(1:end-1);
wolffd@0	27 for i=1:r
wolffd@0	28 for j=1:c
wolffd@0	29 M(i,j) = sum(R(i,j) > cumprob2)+1;
wolffd@0	30 end
wolffd@0	31 end
wolffd@0	32 end
wolffd@0	33
wolffd@0	34
wolffd@0	35 % Slower, even though vectorized
wolffd@0	36 %cumprob = reshape(cumsum([0 prob(1:end-1)]), [1 1 n]);
wolffd@0	37 %M = sum(R(:,:,ones(n,1)) > cumprob(ones(r,1),ones(c,1),:), 3);
wolffd@0	38
wolffd@0	39 % convert using a binning algorithm
wolffd@0	40 %M=bindex(R,cumprob);

Mercurial > hg > camir-aes2014