How to assign probabilities based on certain condition?

Question

I have run into a wall here. So i pull random sample of, lets say 10 elements, and i want to give them probabilities, based on some condition. Lets say that i have this random sample [1, 3, 5, 7, 13, 16, 6, 9, 22, 15] so i want to give each number a probability based on the relative distance from int of 8. Further element is from 8, less the probability. In this case 7 and nine have same, biggest probabilities and 22 have the smallest?

I tried to find some numpy function, but with no success.

Thanks!

Try and look at the answers here: https://stackoverflow.com/questions/3679694/a-weighted-version-of-random-choice — unddoch, Aug 31 '18 at 12:59
Possible duplicate of [A weighted version of random.choice](https://stackoverflow.com/questions/3679694/a-weighted-version-of-random-choice) — Joel, Aug 31 '18 at 13:05

score 1 · Accepted Answer · answered Aug 31 '18 at 13:02

1

You have to do two things. Create the probability, and select according to it. To create it

data = np.array([1, 3, 5, 7, 13, 16, 6, 9, 22, 15])
p = 1 / np.abs(data - 8)

(that's just and idea, you don't specify exactly how you want the probability to change with distance)

For the second part, numpy.choice function (in recent numpy versions) can accept the array p to select with that probability from data

answered Aug 31 '18 at 13:02

blue_note

27,712
9
72
90

Sum of probabilities goes way above 1. I'll just divide every element with sum of probabilitis, that will do the trick. Thanks for quick response! – I.Rankovic Aug 31 '18 at 13:14
@I.Rankovic: yes, just normalize by the sum – blue_note Aug 31 '18 at 13:16

How to assign probabilities based on certain condition?

1 Answers1