View Full Version : [Python] Randomly select list item with a certain probability

February 22nd, 2012, 03:33 PM
I can't seem to get my head round this, although I'm sure it's fairly simple.

I have a list of tuples like this: (c, p) where p is a number between 0 and 1.

I need to return an item from this list, where an item's chance of being selected is equal to p.

So if I had a list like this:
(a, 0.1)
(b, 0.5)
(c, 0.5)

it could return any of them, but b and c are equally likely, and 5 times more likely than a.

Sorry if I haven't explained this very well.

For context, this is part of a genetic algorithm I am writing.

February 22nd, 2012, 03:34 PM

February 22nd, 2012, 03:44 PM

Thanks, that fixed it.


February 22nd, 2012, 03:53 PM
firas@av104151 ~ % cat prob.py
import random

l = [('a', 1), ('b', 5), ('c', 5)]

def randitem(l):

# Make the probabilities add up to 1, preserving ratios
s = sum([b for (a,b) in l])
l2 = []
for (a,b) in l:
l2.append((a, b/s))

r = random.random()
for (a,b) in l2:
if r < b:
return a
r -= b

# test
res = {'a':0, 'b':0, 'c':0}
for i in range(10000):
res[randitem(l)] += 1

print('a: %s (%s%%)' % (res['a'], res['a']/100))
print('b: %s (%s%%)' % (res['b'], res['b']/100))
print('c: %s (%s%%)' % (res['c'], res['c']/100))

firas@av104151 ~ % python3 prob.py
a: 895 (8.95%)
b: 4521 (45.21%)
c: 4584 (45.84%)

February 22nd, 2012, 03:57 PM
Just in case it wasn't clear the first time. ;)

February 22nd, 2012, 04:03 PM
a different approach; i do not normalize the probabilities to 1 (which would probably be the better idea...). this is not well tested!


import random

class RandomChoice(object):

def __init__(self, lst):
lst: the list in the form [ (element, relative_probability) ]

self.lst = lst

self.limits = []
p_sum = 0
for item in lst:
p_sum += item[1]
print self.limits
self.p_max = p_sum

def _get_index(self, rnd):
Given the radnom number rnd (0<=rnd<p_max), return the list index.
last = 0.0
for (i, p) in enumerate(self.limits):
if last <= rnd < p:
return i
last = p

def get(self):
rnd = random.uniform(0, self.p_max)
i = self._get_index(rnd)
return self.lst[i]

# test:

lst = [
('a', 0.1),
('b', 0.5),
('c', 0.5)

rc = RandomChoice(lst)

for i in xrange(100):
print rc.get()

and, yes 'Th' is much more concise!