Let's say I have to generate random number from 1 to 100, however, the probability of each number is not 1/100, but a predefined probability.
How to generate that number? I use Ruby/Python.
the algorithm itself is easy: create an array with all the weights' cumulative sum. After that you generate a [0-1) uniform random number and just binary search that number multiplied with the sum of the weights and the index will be the number you will need to generate.
answered Sep 10 '14 at 7:52
-
2+1 This is the easy way for small, discrete domains. Obviously it doesn't work for continuous (well, as continuous as you can get on a digital computer) and/or analytical distributions. Then you need an analytical solution like the inversion method. – user7043 Sep 10 '14 at 7:57
