np.broadcast_to() for explicit broadcasting in NumPy - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of np.broadcast_to() with 1D to 2D array

What is the output of this code snippet?

NumPy

import numpy as np
arr = np.array([1, 2, 3])
broadcasted = np.broadcast_to(arr, (3, 3))
print(broadcasted)

A

[[1 1 1]
 [2 2 2]
 [3 3 3]]

B

[[1 2 3]
 [2 3 1]
 [3 1 2]]

C

[[1 2 3]
 [1 2 3]
 [1 2 3]]

D [[1 2 3 1 2 3 1 2 3]]

💡 Hint

Broadcasting repeats the original array along new dimensions without copying data.

np.broadcast_to repeats the original 1D array along the new axis to match the target shape (3,3). So each row is the original array.

❓ data_output

intermediate

Shape after broadcasting a scalar

What is the shape of the array after broadcasting this scalar to shape (4, 5)?

NumPy

import numpy as np
scalar = np.array(7)
broadcasted = np.broadcast_to(scalar, (4, 5))
print(broadcasted.shape)

A (1, 1)

B (4,)

C (5, 4)

D (4, 5)

💡 Hint

Broadcasting a scalar repeats it to fill the target shape.

Broadcasting a scalar to (4,5) creates a 2D array with shape (4,5) filled with the scalar value.

🔧 Debug

advanced

Error when broadcasting incompatible shapes

What error does this code raise?

NumPy

import numpy as np
arr = np.array([1, 2, 3])
broadcasted = np.broadcast_to(arr, (2, 2))

A ValueError: shape mismatch: objects cannot be broadcast to a single shape

B TypeError: unsupported operand type(s)

C IndexError: index out of bounds

D No error, output is [[1 2] [3 1]]

💡 Hint

Broadcasting requires the original shape to be compatible with the target shape.

The original array shape (3,) cannot be broadcast to (2,2) because the last dimension sizes differ and are incompatible.

🚀 Application

advanced

Using np.broadcast_to to add bias to a matrix

Given a 2D array of shape (3, 4) and a 1D bias array of shape (4,), which option correctly broadcasts the bias to add it to each row of the matrix?

NumPy

import numpy as np
matrix = np.array([[1, 2, 3, 4],
                   [5, 6, 7, 8],
                   [9, 10, 11, 12]])
bias = np.array([10, 20, 30, 40])

A matrix + np.broadcast_to(bias, (3, 4))

B matrix + np.broadcast_to(bias, (4, 3))

C matrix + np.broadcast_to(bias, (3,))

D matrix + np.broadcast_to(bias, (1, 4))

💡 Hint

The bias must be broadcast to match the matrix shape exactly.

Broadcasting bias to (3,4) repeats it for each row, allowing element-wise addition with the matrix.

🧠 Conceptual

expert

Memory behavior of np.broadcast_to()

Which statement about np.broadcast_to() is TRUE regarding memory usage?

A np.broadcast_to() creates a new array with copied data to match the target shape.

B np.broadcast_to() returns a view that shares the original data without copying.

C np.broadcast_to() always returns a deep copy of the original array.

D np.broadcast_to() modifies the original array in-place to match the new shape.

💡 Hint

Broadcasting tries to avoid copying data for efficiency.

np.broadcast_to() returns a read-only view that shares the original data, avoiding memory copies.