Python Tutorial
NumPy Broadcasting
Broadcasting lets NumPy do math on arrays of different shapes by stretching the smaller one — without copying a full extra array.
Scalar and 1D
import numpy as np
a = np.array([1, 2, 3])
print(a + 10) # [11 12 13]2D plus 1D
m = np.array([[1, 2, 3], [4, 5, 6]])
row = np.array([10, 20, 30])
print(m + row)
# [[11 22 33]
# [14 25 36]]Shapes line up from the right. A dimension of 1 stretches. If sizes differ and neither is 1, you get ValueError.
Column Vector
col = np.array([[10], [20]])
print(m + col)
# [[11 12 13]
# [24 25 26]]📘 Real-World Deep Dive
Knowing <strong>NumPy Broadcasting (NumPy)</strong> well is what turns NumPy from a curiosity into a daily tool — you'll reach for it in nearly every real project.
Real-Life Scenario
An end-to-end usage of NumPy Broadcasting that you'd actually see in a data pipeline or analytics notebook.
Real-Life Example
import numpy as np
matrix = np.ones((3, 3))
vector = np.array([1, 2, 3])
print(matrix + vector) # vector broadcast across rowsExpected Output
(see source)Common mistakes
- NumPy uses 0-based, C-order indexing — the rightmost axis is the *fastest-varying* one. Mixing it with Fortran-order arrays is a common surprise.
np.array([[1,2],[3,4]], dtype=int)is fine, but a ragged Python list produces dtype=object and silently disables vectorisation.- In-place ops (
a *= 2) sometimes break views instead of returning a new array; usenp.multiply(a, 2, out=...)if explicitness matters. - Treating NumPy Broadcasting as a black box without reading the docs — the API has subtle defaults that bite when you scale.
🚀 Performance & Best Practices
- Vectorise: replace Python
forloops with ufuncs; you can expect 10–100× speedups. - Pre-allocate output arrays with
np.emptyinstead of growing them withnp.append. - Keep data in float32 unless you need float64 precision — half the memory, double the cache locality.
- When working with NumPy, prefer vectorised / batched operations over Python loops.
🧪 Try It Yourself
- Reproduce the snippet on a representative slice of your own data.
- Profile the snippet with
cProfileortimeitand find the single biggest improvement. - Generalise the snippet into a small, reusable function you can drop into future projects.