Arrays - random number generating

NumPy - The Basics

2 min read

Published Sep 22 2025

NumPyPython

Random Data in NumPy

NumPy provides several functions in numpy.random to generate arrays with random numbers from different distributions.

Uniform Distribution (0 to 1)

np.random.rand(2, 2)

Generates numbers in the range [0, 1).
Uniformly distributed.
Shape defined by arguments.

Example:

np.random.rand(2, 2)

# → [[0.65 0.12]

# [0.44 0.89]]

Standard Normal Distribution

np.random.randn(8, 2)

Generates values from a standard normal distribution (mean = 0, std = 1).
Useful for simulations and ML initialisation.

Example:

np.random.randn(3, 3)

# → [[ 0.12 -0.89 0.44]

# [ 1.45 0.33 -0.55]

# [-0.23 0.78 -1.09]]

Random Integers

np.random.randint(low=250, high=888, size=(4, 3))

Generates random integers between low (inclusive) and high (exclusive).
Shape defined by size.

Example:

np.random.randint(10, 20, size=(2, 5))

# → [[14 18 12 11 17]

# [19 10 15 12 13]]

Random Sample from [0, 1)

np.random.random(size=(3, 2))

Similar to rand(), but takes only size as a tuple.

Example:

np.random.random((3, 2))

# → [[0.42 0.73]

# [0.89 0.15]

# [0.65 0.22]]

Random Choice

np.random.choice([1, 5, 9], size=(2, 3))

Randomly selects values from a given list/array.
Useful for categorical or discrete sampling.

Example:

np.random.choice([10, 20, 30], size=5)

# → [30 20 10 20 30]

Uniform Distribution

np.random.uniform(low=5, high=15, size=(3, 3))

Random floats between low and high.
Uniform distribution.

Example:

np.random.uniform(5, 10, size=(2, 4))

# → [[5.77 6.12 9.45 7.01]

# [8.22 9.88 5.66 6.99]]

Normal Distribution (custom mean & std)

np.random.normal(loc=50, scale=10, size=(3, 3))

Random floats from a normal distribution with:
- loc = mean
- scale = standard deviation

Example:

np.random.normal(0, 1, size=5)

# → [-0.23 0.87 -1.12 0.44 0.19]

Seeding Random Numbers

Random numbers in NumPy are pseudo-random (deterministic, generated by an algorithm).

np.random.seed(1)

arr = np.random.randint(10, 50, size=5)

print(arr)

# → [37 12 29 37 18] (always the same with seed=1)

np.random.seed(seed) fixes the sequence → ensures reproducibility.
If you set the same seed, you’ll always get the same "random" output.
Useful for debugging, testing, and reproducible research.

How np.random.seed(seed) Works:

All the np.random functions above (rand, randn, randint, random, uniform, normal, choice, etc.) pull numbers from the same underlying random number generator.
When you call np.random.seed(seed), you reset that generator to a fixed state.
That means:
- Using the same seed → you’ll get the same sequence of random numbers.
- Changing the seed → produces a different sequence.
- Resetting to a previous seed → reproduces the exact same values again.