Weighted random selection from array

Question

I would like to randomly select one element from an array, but each element has a known probability of selection.

All chances together (within the array) sums to 1.<

Answer 1

I have found this article to be the most useful at understanding this problem fully. This stackoverflow question may also be what you're looking for.

---

I believe the optimal solution is to use the Alias Method (wikipedia). It requires O(n) time to initialize, O(1) time to make a selection, and O(n) memory.

Here is the algorithm for generating the result of rolling a weighted n-sided die (from here it is trivial to select an element from a length-n array) as take from this article. The author assumes you have functions for rolling a fair die (floor(random() * n)) and flipping a biased coin (random() < p).

Algorithm: Vose's Alias Method

Initialization:

1. Create arrays Alias and Prob, each of size n.
2. Create two worklists, Small and Large.
3. Multiply each probability by n.
4. For each scaled probability p_i:
   1. If p_i < 1, add i to Small.
   2. Otherwise (p_i ≥ 1), add i to Large.
5. While Small and Large are not empty: (Large might be emptied first)
   1. Remove the first element from Small; call it l.
   2. Remove the first element from Large; call it g.
   3. Set Prob[l]=p_l.
   4. Set Alias[l]=g.
   5. Set p_g := (p_g+p_l)−1. (This is a more numerically stable option.)
   6. If p_g<1, add g to Small.
   7. Otherwise (p_g ≥ 1), add g to Large.
6. While Large is not empty:
   1. Remove the first element from Large; call it g.
   2. Set Prob[g] = 1.
7. While Small is not empty: This is only possible due to numerical instability.
   1. Remove the first element from Small; call it l.
   2. Set Prob[l] = 1.

Generation:

1. Generate a fair die roll from an n-sided die; call the side i.
2. Flip a biased coin that comes up heads with probability Prob[i].
3. If the coin comes up "heads," return i.
4. Otherwise, return Alias[i].