Most of the function names in Python can be intuitively connected to the meaning of the function. The NumPy
reshape() function is not an exception. The
reshape() function brings an array into another shape while keeping all the original data.
reshape() function takes an array to be reshaped as a first argument and the new shape tuple as a second argument. It returns a new view on the existing data—if possible—rather than create a full copy of the original array. The returned array behaves like a new object: any change on one view won’t affect any other view.
The Shape Property of a NumPy Array
Before focusing on
shape attribute to find the number of elements along each dimension of this array.
Be careful to remember
The shape attribute always returns a tuple that tells us the length of each dimension. The one-dimensional array is a row vector and its shape is a single value iterable followed by a comma. One-dimensional arrays don’t have rows and columns, so the shape attribute returns a single value tuple.
Let’s look at an example:
import numpy as np #an one-dimensional NumPy array arr = np.arange(10) #print an array print(arr) # [0 1 2 3 4 5 6 7 8 9] print(arr.shape) # (10, )
The code snippet also uses the NumPy
The shape attribute of a two-dimensional array (also called a matrix) gives us a tuple. The shape returns the number of elements along each dimension, which is the number of rows and columns in the two-dimensional array.
# A two-dimensional NumPy array import numpy as np arr = np.array([[1,2,3,4,5], [5,4,3,2,1]]) print(arr.shape) # (2, 5)
The following example is for the shape of three-dimensional arrays.
# A three-dimensional array import numpy as np arr = np.array([ [ [0, 11, 15, 16], [3, 7, 10, 34], [44, 99, 5, 67] ],[ [52, 8, 11, 13], [0, 4, 5, 6], [4, 4, 4, 4] ] ]) print(arr.shape) # (2, 3, 4)
It takes some practice to understand the shape tuple for multidimensional arrays. The dimensions represented by a tuple are read from the outside-in. If you observe the brackets, the outmost bracket is a part of the basic syntax for the whole array. In the shape tuple 2 represents the second set of brackets. If you count them you will see that there are 2 elements in this dimension.
[ [0, 11, 15, 16], [3, 7, 10, 34], [44, 99, 5, 67] ]
[ [52, 8, 11, 13], [0, 4, 5, 6], [4, 4, 4, 4] ]
Each element contains 3 more elements in the second dimension. If you think about nested lists, you can draw the analogy. These elements are:
[0, 11, 15, 16]
[3, 7, 10, 34]
3rd element [
44, 99, 5, 67]
Finally, number 4 represents the number of elements in the third dimension. Those are the innermost elements. For example 0, 11, 15 and 16.
What is the reshape() Function in NumPy
How do we relate NumPy’s
shape attribute to the
numpy.reshape(arr, newshape, order)
arris the array we wish to reshape,
newshapeis an integer for one-dimensional arrays and a tuple of integers multiple-dimensions, and
orderis an optional argument that we won’t be getting into in this guide.
Reshaping an array can be useful when cleaning the data, or if there are some simple element-wise calculations that need to be performed. One of the advantages that NumPy array has over Python list is the ability to perform vectorized operations easier. Moreover, reshaping arrays is common in machine learning. Keep in mind that all the elements in the NumPy array must be of the same type.
Reshape NumPy Array 1D to 2D
Let’s say that we were measuring the outside temperature 3 days in a row, both in Celsius and in Fahrenheit. We recorded our measuring as a one-dimensional vector where all the even indices represent the temperature written in degrees celsius and all the odd indices represent the temperature written in degrees Fahrenheit.
temp = [10, 50, 15, 59, 5, 42]
There are 6 elements recorded in a single row. To reshape the one-dimensional
array to two-dimensional array, we need to pass a tuple with a number
of rows and columns to the reshape function. This tuple will consist of
two numbers, let’s call them m and n, where the first number is the number of rows, and the second number is the number of columns.
One very important note
So, we only have two options for the two-dimensional array, 2 rows, and 3 columns or 3 columns and 2 rows.
import numpy as np temp = [10, 50, 15, 59, 5, 42] temp = np.reshape(temp, (3,2)) print(temp) """ [[10 50] [15 59] [ 5 42]] """
The data hasn’t changed; the same elements are in the same order. They are rearranged in two rows and three columns.
In the section about the shape attribute, we said that the shape of a one-dimensional array is given by a tuple that contains an integer followed by a comma. Then we explained that this vector doesn’t contain rows or columns.
What if we want this vector to have one column and as many rows as there are elements? We can do this
import numpy as np arr = np.arange(10) print(arr.shape) # (10, ) #reshaping this vector arr = np.reshape(arr, (arr.shape, 1)) print(arr.shape) # (10, 1)
Reshape NumPy Array 2D to 1D
Let’s say we are collecting data from a college indoor track meets for the 200-meter dash for women. During the first meet, we record three best times 23.09 seconds, 23.41 seconds, 24.01 seconds. During the second meet, we record three best times 22.55 seconds, 23.05 seconds and 23.09 seconds. We record this in a two-dimensional array. But once we begin analyzing the data we need the results to be in a single row. We do the following to reshape the matrix:
import numpy as np track = np.array([[23.09, 23.41, 24.01], [22.55, 23.05, 23.09]]) track = np.reshape(track, (6,)) print(track) # [23.09 23.41 24.01 22.55 23.05 23.09] print(track.shape) # (6,) print(track.ndim) # 1
NumPy reshape(arr, -1)
Now, we are more likely to have a situation where we have
We can show this in the following example:
import numpy as np track = np.array([[23.09, 23.41, 24.01], [22.55, 23.05, 23.09]]) track = np.reshape(track, -1) print(track) # [23.09 23.41 24.01 22.55 23.05 23.09] print(track.shape) # (6,) print(track.ndim) # 1
Using -1 for the newshape can be useful in the multidimensional arrays. We will get back to it in the next section.
Reshape NumPy Array 3D to 2D
Sometimes the data that we collect will be messy and before we start analyzing it, we need to tidy it up. Let’s say we have a three dimensional NumPy array that looks like this:
data = [[[ 0, 1], [ 2, 3]], [[ 4, 5], [ 6, 7]], [[ 8, 9], [10, 11]], [[12, 13], [14, 15]]]
When we examine the data closer, we can see that it would make more sense to have it stored as a two-dimensional matrix.
We can count the number of “pairs” that we want to have. One way to do this is:
data = np.array(data) data = np.reshape(data, (8,2)) print(data) """ [[ 0 1] [ 2 3] [ 4 5] [ 6 7] [ 8 9] [10 11] [12 13] [14 15]] """
NumPy reshape(-1, m) and reshape(n, -1)
The above method of reshaping three-dimensional array to two-dimensions works if we don’t have a large number of entries. However, if we have thousands of entries, this can be tricky. In this case, we can use -1 for one dimension and if possible the data will be reshaped for us.
Using the example above:
import numpy as np data = [[[ 0, 1], [ 2, 3]], [[ 4, 5], [ 6, 7]], [[ 8, 9], [10, 11]], [[12, 13], [14, 15]]] data = np.array(data) data = np.reshape(data, (-1,2)) print(data) """ [[ 0 1] [ 2 3] [ 4 5] [ 6 7] [ 8 9] [10 11] [12 13] [14 15]] """
np.reshape vs. np.newaxis
When we want to perform operations on arrays, they need to be the compatible size. Element-wise, the size of the arrays needs to be equal in a dimension. But, they don’t have to have the same number of dimensions. If the sizes are not equal, the NumPy throws an error.
When the arrays are different dimensions, one way to add a dimension is using reshape() function. Another way is to
The way it works it “slices” an array by adding a dimension. If we look at the original temperature array from earlier in the guide:
import numpy as np temp = np.array([10, 50, 15, 59, 5, 42]) print(temp.shape) # (6,) temp = temp[np.newaxis, :] print(temp.shape) # (6,1) print(temp) # [[10 50 15 59 5 42]]
This article is contributed by Finxter user Milica Cvetkovic. Thanks for the thorough and detailed work, Milica! 🐍 Milica is also a writer on Medium — check out her Medium Profile.
Where to Go From Here?
A thorough understanding of the NumPy basics is an important part of any data scientist’s education. NumPy is at the heart of many advanced machine learning and data science libraries such as Pandas, TensorFlow, and Scikit-learn.
If you struggle with the NumPy library — fear not! Become a NumPy professional in no time with our new coding textbook “Coffee Break NumPy”. It’s not only a thorough introduction into the NumPy library that will increase your value to the marketplace. It’s also fun to go through the large collection of code puzzles in the book.
While working as a researcher in distributed systems, Dr. Christian Mayer found his love for teaching computer science students.
To help students reach higher levels of Python success, he founded the programming education website Finxter.com. He’s author of the popular programming book Python One-Liners (NoStarch 2020), coauthor of the Coffee Break Python series of self-published books, computer science enthusiast, freelancer, and owner of one of the top 10 largest Python blogs worldwide.
His passions are writing, reading, and coding. But his greatest passion is to serve aspiring coders through Finxter and help them to boost their skills. You can join his free email academy here.