Numpy random number generators

#: standard imports
import numpy as np
# print arrays to 4 decimal places
np.set_printoptions(precision=4, suppress=True)

We often need random numbers, for tests and for taking random samples, and for other things. np.random is a submodule within numpy:

type(np.random)

module

It contains function that will create a random number generator.

# Make a random number generator.
rng = np.random.default_rng()
type(rng)

numpy.random._generator.Generator

This generator is an object that has a set of methods for returning random numbers of various sorts. For example, to return a single random number from the default normal distribution (mean 0, variance 1):

rng.normal()

0.1561313253150664

You can set the mean and variance with the first two input parameters:

# Random number from distribution with mean 15, variance 2
rng.normal(15, 2)

17.283947497270667

To return a 8 by 5 array of random numbers from the same distribution:

rng.normal(15, 2, size=(8, 5))

array([[14.5994,  9.7962, 16.2723, 16.752 , 17.5274],
       [14.8919, 18.1976, 11.4291, 12.2943, 17.4066],
       [14.649 ,  9.7467, 15.0475, 16.0897, 10.9309],
       [14.8456, 11.4277, 13.2358, 19.0015, 17.5183],
       [17.4964, 17.6885,  8.1901, 13.7035, 15.9571],
       [14.7317, 17.8072, 17.5164, 15.8536, 10.639 ],
       [18.3061, 12.8098, 15.2429, 13.2912, 13.2401],
       [13.8735, 15.6828, 13.8814, 16.8983, 14.55  ]])

A 5 by 3 array of random numbers from the standard normal distribution with mean 1 and variance 1:

rng.normal(size=(5, 3))

array([[ 0.2159,  0.7549, -0.6959],
       [ 0.4839, -0.0873, -1.5388],
       [ 0.3123, -0.0434,  1.2928],
       [-0.3932, -0.5191,  1.1985],
       [ 0.5762, -0.4437, -0.7948]])

Making random numbers predictable

Sometimes you want to make sure that the random numbers are predictable, in that you will always get the same set of random numbers from a series of calls to the rng methods. You can achieve this by giving the random number generator a seed when you create it. This is an integer that sets the random number generator into a predictable state, such that it will always return the same sequence of random numbers from this point:

# Set the state of the random number generator on creation.
new_rng = np.random.default_rng(seed=42)
# One set of random numbers
first_random_arr = new_rng.normal(size=(4, 2))
first_random_arr

array([[ 0.3047, -1.04  ],
       [ 0.7505,  0.9406],
       [-1.951 , -1.3022],
       [ 0.1278, -0.3162]])

# Another set
second_random_arr = new_rng.normal(size=(4, 2))
second_random_arr

array([[-0.0168, -0.853 ],
       [ 0.8794,  0.7778],
       [ 0.066 ,  1.1272],
       [ 0.4675, -0.8593]])

# Make another random number generator with the same seed.
new_rng2 = np.random.default_rng(seed=42)
# The same as "first_random_arr" above.
new_rng2.normal(size=(4, 2))

array([[ 0.3047, -1.04  ],
       [ 0.7505,  0.9406],
       [-1.951 , -1.3022],
       [ 0.1278, -0.3162]])

# The same as "second_random_arr" above.
new_rng2.normal(size=(4, 2))

array([[-0.0168, -0.853 ],
       [ 0.8794,  0.7778],
       [ 0.066 ,  1.1272],
       [ 0.4675, -0.8593]])