How to flatten only some Dimensions of a NumPy array

# Table of Contents

# How to flatten only some Dimensions of a NumPy array

Use the numpy.reshape() method to flatten only some dimensions of a NumPy array.

The method will flatten the array, giving it a new shape, without changing its data.

main.py

Copied!
import numpy as np

arr = np.zeros((2, 4, 2))
print(arr)

print('-' * 50)

new_arr = arr.reshape(8, 2)
print(new_arr)

The code for this article is available on GitHub

Running the code sample produces the following output.

shell

Copied!
[[[0. 0.]
  [0. 0.]
  [0. 0.]
  [0. 0.]]

 [[0. 0.]
  [0. 0.]
  [0. 0.]
  [0. 0.]]]
--------------------------------------------------
[[0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]]

flatten only some dimensions of numpy array

The numpy.reshape() method gives a new shape to an array without changing its data.

The only arguments we passed to the reshape() method are 2 integers that represent the new shape.

The new shape should be compatible with the original shape.

main.py

Copied!
import numpy as np

arr = np.zeros((2, 4, 2))

new_arr = arr.reshape(8, 2)

# [[0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]]
print(new_arr)

The code for this article is available on GitHub

# Using `-1` as a shape dimension when flattening the array

One shape dimension can be -1.

When a shape dimension is -1, its value is inferred from the length of the array and remaining dimensions.

main.py

Copied!
import numpy as np

arr = np.zeros((2, 4, 2))

new_arr = arr.reshape(-1, arr.shape[-1])

# [[0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]
#  [0. 0.]]
print(new_arr)

one shape dimension can be minus one

The code for this article is available on GitHub

The last shape dimension in the example is 2, so the first shape dimension (-1) is inferred to be 8.

main.py

Copied!
import numpy as np

arr = np.zeros((2, 4, 2))

print(arr.shape[-1])  # 👉️ 2

This is calculated by dividing the total size of the array (16) by the product of all other specified dimensions (2) = 8.

main.py

Copied!
import numpy as np

arr = np.zeros((2, 4, 2))

print(arr.size)  # 👉️ 16

print(arr.shape[-1])  # 👉️ 2

print(arr.size // arr.shape[-1])  # 👉️ 8

The code sample flattens all but the last dimension.

You can use a similar approach to flatten all but the last two dimensions.

main.py

Copied!
import numpy as np

arr = np.zeros((2, 4, 2, 4))

new_arr = arr.reshape(-1, *arr.shape[-2:])
print(new_arr)

Running the code sample produces the following output.

shell

Copied!
[[[0. 0. 0. 0.]
  [0. 0. 0. 0.]]

 [[0. 0. 0. 0.]
  [0. 0. 0. 0.]]

 [[0. 0. 0. 0.]
  [0. 0. 0. 0.]]

 [[0. 0. 0. 0.]
  [0. 0. 0. 0.]]

 [[0. 0. 0. 0.]
  [0. 0. 0. 0.]]

 [[0. 0. 0. 0.]
  [0. 0. 0. 0.]]

 [[0. 0. 0. 0.]
  [0. 0. 0. 0.]]

 [[0. 0. 0. 0.]
  [0. 0. 0. 0.]]]

You should always make sure that the new shape is compatible with the original shape.

# Flatten only some Dimensions of a NumPy array using `numpy.vstack()`

You can also use the numpy.vstack() method to flatten only some dimensions of a NumPy array.

The method stacks the arrays in a sequence vertically (row-wise).

main.py

Copied!
import numpy as np

arr = np.zeros((2, 4, 2))
print(arr)

print('-' * 50)

new_arr = np.vstack(arr)
print(new_arr)

print('-' * 50)

print(new_arr.shape)

The code for this article is available on GitHub

Running the code sample produces the following output.

shell

Copied!
[[[0. 0.]
  [0. 0.]
  [0. 0.]
  [0. 0.]]

 [[0. 0.]
  [0. 0.]
  [0. 0.]
  [0. 0.]]]
--------------------------------------------------
[[0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]
 [0. 0.]]
--------------------------------------------------
(8, 2)

flatten only some dimensions of numpy array using numpy vstack

The numpy.vstack() method is equivalent to concatenating along the first axis after 1-D arrays of shape (N,) have been reshaped to (1, N).

# Additional Resources

You can learn more about the related topics by checking out the following tutorials:

I wrote a book in which I share everything I know about how to become a better, more efficient programmer.

You can use the search field on my Home Page to filter through all of my articles.

How to flatten only some Dimensions of a NumPy array

# Table of Contents