Random Arrays

np.random builds arrays of random values for tests, simulations, and sample data. Pass size as a number or a shape tuple. Set a seed with np.random.seed(n) when you want the same values on every run of the same machine.

randint

randint(high) picks one whole number from 0 up to high - 1. randint(low, high, size) fills a shape. The high value is never picked, so a six-sided die is randint(1, 7).

import numpy as np

np.random.seed(7)
rolls = np.random.randint(1, 7, size=10)
print(rolls)
print(rolls.min() >= 1, rolls.max() <= 6)
print(np.random.randint(1, 100, size=(2, 3)).shape)

Output from one run (random values change between runs and machines):

[5 2 4 4 5 2 1 2 3 3]
True True
(2, 3)

rand and uniform

rand gives floats from 0 up to, but not including, 1. Pass the dimensions as separate numbers. uniform(low, high, size) gives floats in your own range.

import numpy as np

np.random.seed(7)
print(np.round(np.random.rand(3), 3))
print(np.random.rand(2, 4).shape)
temps = np.random.uniform(18, 26, size=4)
print(np.round(temps, 1))

Output from one run (random values change between runs and machines):

[0.076 0.78  0.438]
(2, 4)
[24.4 21.  18.5 20.3]

randn and normal

randn draws from the standard normal distribution: mean 0 and standard deviation 1. normal(loc, scale, size) sets the mean with loc and the spread with scale. With many samples, the measured mean and spread land close to what you asked for.

import numpy as np

np.random.seed(7)
print(np.round(np.random.randn(4), 3))
heights = np.random.normal(loc=170, scale=8, size=10000)
print(round(heights.mean()), round(heights.std()))

Output from one run (random values change between runs and machines):

[ 1.691 -0.466  0.033  0.408]
170 8

choice, shuffle, and permutation

choice(values, size) picks items, with repeats allowed. shuffle reorders an array in place and returns None. On a 2-D array it moves whole rows. permutation returns a reordered copy and leaves the original unchanged.

import numpy as np

np.random.seed(7)
print(np.random.choice(["red", "green", "blue"], size=5))

cards = np.arange(1, 6)
np.random.shuffle(cards)
print(cards)

original = np.arange(1, 6)
mixed = np.random.permutation(original)
print(mixed, original)

Output from one run (random values change between runs and machines):

['red' 'green' 'blue' 'red' 'green']
[4 5 3 2 1]
[2 4 5 1 3] [1 2 3 4 5]

Newer code often creates a generator with rng = np.random.default_rng(7) and calls rng.integers(1, 7, size=5) or rng.normal(170, 8, size=5). The ideas are the same.

Key takeaways

• randint(low, high) never returns high

• rand and uniform give floats. randn and normal give a bell curve

• shuffle changes the array and returns None

• permutation returns a new order and keeps the original

• Next: Indexing

• Previous: Creating Arrays