What is broadcasting?

Broadcasting is NumPy’s ability to perform operations on arrays with different shapes by automatically expanding (virtually) smaller arrays.

This is a major reason NumPy code is concise and fast. Nothing is actually copied in memory — NumPy just repeats the smaller array’s values as if it were stretched to fit, without allocating that extra memory.

Simple example: add scalar

import numpy as np
 
arr = np.array([1, 2, 3])
print(arr + 10)  # [11 12 13]

import numpy as np
 
arr = np.array([1, 2, 3])
print(arr + 10)  # [11 12 13]

Here, the scalar 1010 is broadcast to match the shape (3,)(3,).

Example: add vector to matrix

import numpy as np
 
mat = np.array([
    [1, 2, 3],
    [4, 5, 6]
])
vec = np.array([10, 20, 30])
 
print(mat + vec)

import numpy as np
 
mat = np.array([
    [1, 2, 3],
    [4, 5, 6]
])
vec = np.array([10, 20, 30])
 
print(mat + vec)

vecvec (shape (3,)(3,)) broadcasts across each row.

Broadcasting rules (must know)

When operating on two arrays, NumPy compares shapes from the trailing dimension (rightmost first) and works backward.

Two dimensions are compatible when:

1. They are equal, OR
2. One of them is 11

If dimensions are incompatible → broadcasting error.

Example: column vector + matrix

import numpy as np
 
mat = np.array([
    [1, 2, 3],
    [4, 5, 6]
])
col = np.array([100, 200]).reshape(2, 1)
 
print(mat + col)

import numpy as np
 
mat = np.array([
    [1, 2, 3],
    [4, 5, 6]
])
col = np.array([100, 200]).reshape(2, 1)
 
print(mat + col)

colcol has shape (2, 1)(2, 1) and broadcasts across columns.

Common broadcasting error

import numpy as np
 
mat = np.zeros((2, 3))
vec = np.array([1, 2])
 
# mat + vec -> ValueError (shapes (2,3) and (2,) not compatible)

import numpy as np
 
mat = np.zeros((2, 3))
vec = np.array([1, 2])
 
# mat + vec -> ValueError (shapes (2,3) and (2,) not compatible)

Fix by reshaping vecvec to a column vector if that’s what you intend:

vec = vec.reshape(2, 1)
print(mat + vec)

vec = vec.reshape(2, 1)
print(mat + vec)

  flowchart TD
  A["Compare shapes,
trailing axis first"] --> B{"Axis sizes equal
or one is 1?"}
  B -- "Yes, for every axis" --> C["Broadcast succeeds
virtually stretch the smaller array"]
  B -- "No, some axis mismatches" --> D["ValueError:
shapes not aligned"]

Why broadcasting is useful in analytics

• Normalize columns: X / X.max(axis=0)X / X.max(axis=0)
• Center data: X - X.mean(axis=0)X - X.mean(axis=0)
• Apply weights: X * weightsX * weights

Next

Continue to: NumPy Arithmetic Operations for element-wise math and matrix operations.

🧪 Try It Yourself

Exercise 1 – Broadcast a Scalar

Exercise 2 – Broadcast a Row Vector Across a Matrix

Exercise 3 – Fix a Shape Mismatch