Returning a value at random based on a probability weights

Question

Possible Duplicate:
A weighted version of random.choice

Let's say for simplicity a function takes in 4 inputs: 2 names and their respective 'bias/weights', how can I write a function such that it returns either a or b at random but still considers those weights into the random outcome.

a = 'x'
b = 'y'
p_a = 10
p_b = 90

def myRand(a, p_a, b, p_b):
    return 'x' 10% of the time, 'y' 90% of time

What I've done so far

import random

def myRand(a, p_a, b, p_b):
    probs = [a]*p_a + [b]*p_b
    return random.choice(probs)

Can someone point why this is an incorrect or not the best answer? My justification is that I'm assuming every element has the same probability of being picked so the outcome should still favor 10:90. Or maybe we can shuffle the array before using random.choice() ?

Is there a better way to do this? Maybe I'm missing something obvious here or is this correct?

Thanks

What you're doing will work, with two options and integer weights that aren't too large, but the accepted answer in the linked question is more general and should perform better. — DSM, Aug 23 '12 at 17:21

Lanaru · Answer 1 · 2012-08-23T17:23:47.433

2

This would accomplish what you're trying to do:

#assumes p_a and p_b are both positive numbers that sum to 100
def myRand(a, p_a, b, p_b):
   return a if random.uniform(0,100) < p_a else b

Note that p_b becomes unnecessary in your special case of having only 2 weights, since p_b == 100-p_a.

edited Aug 23 '12 at 17:23

answered Aug 23 '12 at 17:15

Lanaru

9,421
7
38
64

chucksmash · Accepted Answer · 2012-08-23T18:54:41.197

1

I've modified the function to accept an arbitrary number of inputs and weighted probabilities so if later on you decide you want to use more than two inputs, you can.

import random

def myRand(i, w):
    r = random.uniform(0, sum(w))

    # loop through a list of inputs and max cutoff values, returning
    # the first value for which the random num r is less than the cutoff value
    for n,v in map(None, i,[sum(w[:x+1]) for x in range(len(w))]):
        if r < v:
            return n

Example usage:

inputs = ['e', 'f', 'g', 'h']
weights = [10, 30, 50, 10]

print myRand(inputs, weights)

edited Aug 23 '12 at 18:54

answered Aug 23 '12 at 18:42

chucksmash

5,777
1
32
41

Traceback (most recent call last): File "randomwalk.py", line 48, in P = random_win(P, win_rate, win_amount, loss_amount) File "randomwalk.py", line 41, in random_win return myRand([P+win_amount, P-loss_amount], [win_rate, 1-win_rate]) File "randomwalk.py", line 35, in myRand for n,v in map(None, i,[sum(w[:x+1]) for x in range(len(w))]): TypeError: 'NoneType' object is not callable – Aditya Mittal Dec 17 '17 at 17:39

score 0 · Answer 3 · answered Aug 23 '12 at 17:16

0

a = 'x';b = 'y';p_a = 10;p_b = 90

ratio = p_a+pb = 100

generate a random number between 0 and 100 and if the number is less than 10 use a=>x else use b=>y

answered Aug 23 '12 at 17:16

chicharito

1,047
3
12
41

as what i am thing , correct ans – user1085195 Aug 23 '12 at 17:17

Returning a value at random based on a probability weights

3 Answers3

Linked