np.broadcast_to() for explicit broadcasting in NumPy

Introduction

We use np.broadcast_to() to stretch a smaller array to a bigger shape without copying data. This helps when you want to do math with arrays of different sizes.

You have a small array and want to match the shape of a bigger array for calculations.

You want to repeat data logically without making a full copy in memory.

You need to prepare arrays for element-wise operations in machine learning.

You want to avoid writing loops by using array broadcasting explicitly.

You want to understand how NumPy handles arrays of different shapes.

Syntax

NumPy

np.broadcast_to(array, shape, subok=False)

array is the input array you want to broadcast.

shape is the new shape you want to broadcast to. It must be compatible.

Examples

This repeats the 1D array x into a 3x3 array by broadcasting.

NumPy

import numpy as np

x = np.array([1, 2, 3])
y = np.broadcast_to(x, (3, 3))
print(y)

Broadcast a 3x1 array to 3x4 by repeating the single column.

NumPy

a = np.array([[1], [2], [3]])
b = np.broadcast_to(a, (3, 4))
print(b)

Sample Program

This code shows how a 1D array of length 2 is broadcasted to a 3x2 array. The values are repeated logically without copying.

NumPy

import numpy as np

# Original small array
arr = np.array([10, 20])

# Broadcast to a bigger shape
broadcasted_arr = np.broadcast_to(arr, (3, 2))

print("Original array shape:", arr.shape)
print("Broadcasted array shape:", broadcasted_arr.shape)
print("Broadcasted array:")
print(broadcasted_arr)

Important Notes

The new shape must be compatible with the original array shape, or you get an error.

np.broadcast_to() does not copy data, so it is memory efficient.

Broadcasted arrays are read-only to avoid accidental changes.

Summary

np.broadcast_to() stretches arrays to bigger shapes for easy math.

It repeats data logically without copying, saving memory.

Use it when you want to prepare arrays for element-wise operations.