duncan smith wrote:
Hello,
Just checking to see if anyone has attacked this problem before
for cases where the population size is unfeasibly large.
The fastest way I know of is to create a list of cumulative frequencies, then generate uniformly distributed numbers and use a binary search to find where they fall in the list. That's O(log n) per sample in the size of the list once it's been set up. -- Greg -- https://mail.python.org/mailman/listinfo/python-list
