Working with NumPy Arrays

Site Admin · 11 Sep 2026 · 10 views

Working with NumPy Arrays

NumPy is the foundation under almost every Python ML library. Its ndarray stores numbers in contiguous memory and runs vectorized operations without Python loops. That speed difference is what makes real datasets feasible.

Create Arrays

import numpy as np

a = np.array([1, 2, 3])
b = np.ones((2, 3))
c = np.zeros(4)
d = np.arange(0, 10, 2)
print(a, b, c, d)

Vectorized Math

Operations apply element by element without explicit loops:

x = np.array([1, 2, 3, 4])
y = x * 2
z = x + y
print(x.mean())
print(x.sum())
print(x ** 2)

Why does this matter? A loop over a million Python floats is slow. The same operation over a NumPy array runs in compiled C. Always prefer vectorized expressions such as x * 2 over manual iteration.

Shapes and Reshaping

Understand shape, size, and dtype. Shape mismatches are the most common NumPy error, and a read of the shape usually shows the fix.

m = np.arange(12).reshape(3, 4)
print(m.shape)
print(m[1, 2])
print(m.sum(axis=0))

axis=0 collapses rows; axis=1 collapses columns. Practice on a small matrix until the two feel natural.

Indexing and Slicing

NumPy indexing uses the same brackets as lists but supports fancy indexing with arrays of positions and boolean masks, which you will use constantly for filtering data.