How to generate highly uniform random number in python?

Question

In somehow I need an random function that can generate highly uniform numbers, such as, when I say [5,5], and I run this function for 10 times, I would like to have 5 number that are less than 0.5 and other 5 numbers are greater than 0.5.

Every run should be separated, like:

for i in range(0,10):
   rnd=random_function([6,3,1])
   print rnd

After having run this function for ten times, then I have 6 numbers are less than 0.6, 3 numbers are between 0.6 and 0.9, and 1 number are between 0.9 to 1.

Does anyone know how can I do that? Because the random function inside python is not that uniform.... Thanks!!

Answer 1

What you are asking for is not random; indeed, it is fairly structured. The point of random data is that it isn't structured: you can't say anything about what the values will be, only what they will be on average.

You can still do what you want, though: just specify the ranges you want your numbers to fall into.

>>> def random_function(*args):
...     n = sum(args)
...     bottom = 0
...     for m in args:
...             for _ in range(m):
...                     yield random.uniform(bottom, bottom + m/n)
...             bottom += m/n
...
>>> list(random_function(6,3,1))
[0.1849778317803791, 0.2779140519434712, 0.08380168147928498, 0.5477412922676888
, 0.5158697440011519, 0.5535466918038039, 0.8046993690361345, 0.714614514522802,
 0.7102988180048052, 0.9608096335125095]
>>> list(random_function(6,3,1))
[0.29313403848522546, 0.5543469551407342, 0.14208842652528347, 0.344464024804118
7, 0.3168266508239002, 0.5956620829410604, 0.673021479097414, 0.7141779120687652
, 0.7783099010964684, 0.9103924993423906]

---

Explanation:

Given (6,3,1), work out their sum (10). We want six numbers between 0 and 6/10, three numbers between 6/10 and 6/10 + 3/10, and one number between 6/10 + 3/10 and 6/10 + 3/10 + 1/10. We can generate these by calling

random.uniform(0, 6/10) six times,
random.uniform(6/10, 9/10) three times, and
random.uniform(9/10, 1) once.

Answer 2

Isn't it the essense of random that it's not uniform? If you want to generate 6 random numbers and you want half of them to be under 0.5 and the other half to be over 0.5, you should generate 3 numbers between 0 and 0.5 and generate 3 between 0.5 and 1.

Answer 3

Well, that's kind of defeating the purpose of "random" numbers, but Python does support this.

As part of the random module (import random), you can use random.uniform(a,b) instead of random.random() where a to b is the range in floating point.

More info in Python's documentation.

Answer 4

It would help to know the context of your question, but it appears that what you may be looking for are low-discrepancy sequences rather than the usual pseudo-random numbers.

More specifically, check if Halton sequence gives you desired results. Python implementation is available.