NumPy - Finding Unique Rows

---

---

Finding Unique Rows in NumPy Array

In NumPy, arrays can contain multiple rows of data, and sometimes you might want to identify rows that are unique, meaning they appear only once in the array. Finding unique rows involves determining which rows are distinct from others based on their content.

In NumPy, we can achieve this using the unique() function.

Using union1d() Function

The np.unique() function is commonly used to find unique elements in an array. When applied with the axis parameter, it can be used to find unique rows. Following is the syntax −

numpy.unique(a, axis=None, return_index=False, return_inverse=False, return_counts=False)

Where,

• a − It is the input array.
• axis − It is the axis along which to find unique values. Set to 0 for rows.
• return_index − It determines whether to return the indices of the first occurrences.
• return_inverse − It determines whether to return the indices that can reconstruct the array.
• return_counts − It determines whether to return the counts of unique values.

Example: Finding Unique Elements in a 1D Array

The simplest use of np.unique() function is to find unique elements in a one-dimensional array −

import numpy as np

# Define a 1D array with duplicate values
array = np.array([1, 2, 2, 3, 4, 4, 5])

# Find unique elements
unique_elements = np.unique(array)

print("Unique Elements:\n", unique_elements)

Following is the output obtained −

Unique Elements:
[1 2 3 4 5]

Example: Unique Rows in a 2D Array

In the following example, we are using the unique() function to retrieve the unique rows in a 2D array, removing any duplicate rows −

import numpy as np

# Define an array with duplicate rows
array = np.array([
    [1, 2, 3],
    [4, 5, 6],
    [1, 2, 3],
    [7, 8, 9]
])

# Find unique rows
unique_rows = np.unique(array, axis=0)

print("Unique Rows:\n", unique_rows)

This will produce the following result −

Unique Rows:
[[1 2 3]
 [4 5 6]
 [7 8 9]]

Finding Unique Rows with Indexes

We can find the indices of the unique rows in the original array in NumPy by setting the return_index parameter to True in the unique() function.

Example

In this example, we are finding unique rows and their indices using the unique() function −

import numpy as np

# Define an array with duplicate rows
array = np.array([
    [1, 2, 3],
    [4, 5, 6],
    [1, 2, 3],
    [7, 8, 9]
])

# Find unique rows and their indices
unique_rows, indices = np.unique(array, axis=0, return_index=True)

print("Unique Rows:\n", unique_rows)
print("Indices of Unique Rows:\n", indices)

Following is the output of the above code −

Unique Rows:
[[1 2 3]
 [4 5 6]
 [7 8 9]]
Indices of Unique Rows:
[0 1 3]

Reconstructing the Original Array

If you need to reconstruct the original array from the unique rows, you can use the indices returned by np.unique() function with the return_inverse parameter set to True. The inverse indices can be used to map back to the original data from the unique values.

Example

In this example, we are identifying unique rows in a NumPy array and their original indices using the unique() function. We then reconstruct the array using these indices to verify that the unique rows match the original array without duplicates −

import numpy as np

# Define an array with duplicate rows
array = np.array([
    [1, 2, 3],
    [4, 5, 6],
    [1, 2, 3],
    [7, 8, 9]
])

# Find unique rows and their indices
unique_rows, indices = np.unique(array, axis=0, return_index=True)

# Reconstruct the original array using the indices
reconstructed_array = array[np.sort(indices)]

print("Reconstructed Array:\n", reconstructed_array)

The output obtained is as shown below −

Reconstructed Array:
[[1 2 3]
 [4 5 6]
 [7 8 9]]

Counting Unique Rows

In addition to finding unique rows, you might want to count how many times each unique row appears in the array. In NumPy, you can achieve this by setting the return_counts parameter to True in the unique() function.

This is particularly useful when working with multi-dimensional arrays where each row represents a record or observation.

Example

In the following example, we retrieve the count of each unique row in the original array using the unique() function −

import numpy as np

# Define an array with duplicate rows
array = np.array([
    [1, 2, 3],
    [4, 5, 6],
    [1, 2, 3],
    [7, 8, 9]
])

# Find unique rows and their counts
unique_rows, counts = np.unique(array, axis=0, return_counts=True)

print("Unique Rows:\n", unique_rows)
print("Counts of Each Row:\n", counts)

After executing the above code, we get the following output −

Unique Rows:
[[1 2 3]
 [4 5 6]
 [7 8 9]]
Counts of Each Row:
[2 1 1]

Multi-dimensional Arrays

For multi-dimensional arrays, you can use np.unique() function to find unique rows by setting the axis parameter to 0. To handle unique values across all dimensions, you can use the default settings.

Example

In the example below, we flatten the 3D array into 2D and then find unique rows using the unique() function −

import numpy as np

# Define a 3D array
array = np.array([
    [[1, 2], [3, 4]],
    [[1, 2], [5, 6]],
    [[1, 2], [3, 4]]
])

# Flatten the 3D array to 2D for uniqueness check
array_2d = array.reshape(-1, array.shape[-1])

# Find unique rows in the flattened array
unique_rows = np.unique(array_2d, axis=0)

print("Unique Rows in 3D Array:\n", unique_rows)

The result produced is as follows −

Unique Rows in 3D Array:
[[1 2]
 [3 4]
 [5 6]]