wolffd@0: function M = sample_discrete(prob, r, c)
wolffd@0: % SAMPLE_DISCRETE Like the built in 'rand', except we draw from a non-uniform discrete distrib.
wolffd@0: % M = sample_discrete(prob, r, c)
wolffd@0: %
wolffd@0: % Example: sample_discrete([0.8 0.2], 1, 10) generates a row vector of 10 random integers from {1,2},
wolffd@0: % where the prob. of being 1 is 0.8 and the prob of being 2 is 0.2.
wolffd@0: 
wolffd@0: n = length(prob);
wolffd@0: 
wolffd@0: if nargin == 1
wolffd@0:   r = 1; c = 1;
wolffd@0: elseif nargin == 2
wolffd@0:   c == r;
wolffd@0: end
wolffd@0: 
wolffd@0: R = rand(r, c);
wolffd@0: M = ones(r, c);
wolffd@0: cumprob = cumsum(prob(:));
wolffd@0: 
wolffd@0: if n < r*c
wolffd@0:   for i = 1:n-1
wolffd@0:     M = M + (R > cumprob(i));
wolffd@0:   end
wolffd@0: else
wolffd@0:   % loop over the smaller index - can be much faster if length(prob) >> r*c
wolffd@0:   cumprob2 = cumprob(1:end-1);
wolffd@0:   for i=1:r
wolffd@0:     for j=1:c
wolffd@0:       M(i,j) = sum(R(i,j) > cumprob2)+1;
wolffd@0:     end
wolffd@0:   end
wolffd@0: end
wolffd@0: 
wolffd@0: 
wolffd@0: % Slower, even though vectorized
wolffd@0: %cumprob = reshape(cumsum([0 prob(1:end-1)]), [1 1 n]);
wolffd@0: %M = sum(R(:,:,ones(n,1)) > cumprob(ones(r,1),ones(c,1),:), 3);
wolffd@0: 
wolffd@0: % convert using a binning algorithm
wolffd@0: %M=bindex(R,cumprob);