34 lines
870 B
Matlab
34 lines
870 B
Matlab
|
% Weighted random sampling.
|
||
|
|
%
|
||
|
|
% INPUTs: number of draws from a discrete distribution (n)
|
||
|
|
% possible values to pick from, (P)
|
||
|
|
% set of normalized weights/probabilities, (W)
|
||
|
|
% OUTPUTs: s - set of n numbers drawn from P
|
||
|
|
% according to the weights in W
|
||
|
|
%
|
||
|
|
% GB: last updated, Oct 31 2012
|
||
|
|
|
||
|
|
|
||
|
|
function s = weightedRandomSample(n,P,W)
|
||
|
|
|
||
|
|
s = [];
|
||
|
|
|
||
|
|
if abs(sum(W)-1)>10**(-8);
|
||
|
|
printf('The probabilities do not sum up to 1.\n');
|
||
|
|
s = 'probabilities do not sum up to 1';
|
||
|
|
return;
|
||
|
|
end
|
||
|
|
|
||
|
|
% divide the unit interval into |P| segments each with length W_i
|
||
|
|
unit = [0,W(1)];
|
||
|
|
for w=2:length(W)
|
||
|
|
unit = [unit W(w)+unit(length(unit))];
|
||
|
|
end
|
||
|
|
|
||
|
|
% draw a random number in the unit interval uniformly - where does it fall?
|
||
|
|
while length(s)<n
|
||
|
|
|
||
|
|
lb = max(find(unit<rand)); % rand is in [unit(lb), unit(lb+1)]
|
||
|
|
s = [s P(lb)]; % pick P(lb)
|
||
|
|
|
||
|
|
end
|