How to generate random arrays with NumPy

Random arrays let NumPy code simulate measurements, build repeatable tests, and create synthetic inputs without hand-writing sample data. One generator object can draw floating-point values, bounded integers, and distribution-based samples while returning normal ndarray objects.

np.random.default_rng() creates a Generator, which is the modern NumPy random API. Its methods use size for output shape, so a tuple such as (2, 3) creates a two-row, three-column array.

Use an explicit seed for documentation, tests, and experiments that need the same values on repeated runs. Leave the seed out when each run should draw a new sequence from the operating system's entropy source.

1. Create a script that draws floating-point, integer, and normally distributed arrays from one generator.

   random-array-generate.py

   import numpy as np
    
   rng = np.random.default_rng(seed=2026)
    
   uniform = rng.random(size=(2, 3))
   integers = rng.integers(low=1, high=10, size=(2, 3))
   normal = rng.normal(loc=100, scale=5, size=6)
    
   print("uniform shape:", uniform.shape)
   print(np.round(uniform, 3))
   print("integers shape:", integers.shape)
   print(integers)
   print("normal mean:", round(float(normal.mean()), 2))
   print("normal std:", round(float(normal.std(ddof=1)), 2))
    
   shape_checks = (
       uniform.shape == (2, 3)
       and integers.shape == (2, 3)
       and normal.shape == (6,)
   )
   integer_bounds = ((integers >= 1) & (integers < 10)).all()
    
   print("shape checks passed:", shape_checks)
   print("integer bounds passed:", bool(integer_bounds))

   rng.random(size=(2, 3)) returns values in [0.0, 1.0). rng.integers(low=1, high=10, size=(2, 3)) returns integers from 1 through 9 because high is excluded by default.

2. Run the script and verify the generated shapes and integer bounds.

   $ python3 random-array-generate.py
   uniform shape: (2, 3)
   [[0.179 0.64  0.467]
    [0.371 0.355 0.791]]
   integers shape: (2, 3)
   [[7 9 7]
    [2 8 6]]
   normal mean: 100.85
   normal std: 1.87
   shape checks passed: True
   integer bounds passed: True

   The pass lines confirm that size produced the requested shapes and that integers() kept every value inside the configured low-inclusive, high-exclusive range. Remove seed=2026 when repeatable output is not required.