Optimizing Dot product calculation of large NumPy arrays

NumPy: Performance Optimization Exercise-3 with Solution

Write a function to compute the dot product of two large NumPy arrays using a nested for loop. Then, optimize it using NumPy's dot() function.

Sample Solution:

Python Code:

import numpy as np
# Generate two large 2D NumPy arrays with shape (1000, 1000)
array1 = np.random.rand(1000, 1000)
array2 = np.random.rand(1000, 1000)

# Function to compute the dot product using nested for loops
def dot_product_using_loops(arr1, arr2):
    result = np.zeros((arr1.shape[0], arr2.shape[1]))
    for i in range(arr1.shape[0]):
        for j in range(arr2.shape[1]):
            for k in range(arr1.shape[1]):
                result[i, j] += arr1[i, k] * arr2[k, j]
    return result

# Compute the dot product using the nested for loops
dot_product_loops = dot_product_using_loops(array1, array2)
print("Dot product using nested for loops (first 5x5 block):\n", dot_product_loops[:5, :5])

# Optimize the dot product computation using NumPy's dot() function
dot_product_numpy = np.dot(array1, array2)
print("Dot product using NumPy's dot() function (first 5x5 block):\n", dot_product_numpy[:5, :5])

Output:

Dot product using nested for loops (first 5x5 block):
 [[24.22324898 25.99396249 24.81243363 26.47654967 21.87978998]
 [23.34320941 27.11995317 25.42488123 26.39399097 24.27911587]
 [26.08843058 27.77525101 25.13556487 27.49009703 24.78556125]
 [24.07227323 25.87107398 23.25500579 25.07011204 23.13590395]
 [24.31020899 28.82889983 26.5881906  26.41558584 23.82757658]]
Dot product using NumPy's dot() function (first 5x5 block):
 [[24.22324898 25.99396249 24.81243363 26.47654967 21.87978998]
 [23.34320941 27.11995317 25.42488123 26.39399097 24.27911587]
 [26.08843058 27.77525101 25.13556487 27.49009703 24.78556125]
 [24.07227323 25.87107398 23.25500579 25.07011204 23.13590395]
 [24.31020899 28.82889983 26.5881906  26.41558584 23.82757658]]

Explanation:

• Generate two large arrays: Two large 2D NumPy arrays, each with shape (1000, 1000), are created using np.random.rand().
• Function with nested for loops: A function dot_product_using_loops computes the dot product using nested for loops.
• Calculate dot product with loops: The dot product of the arrays is calculated using the nested for loops, and the first 5x5 block of the result is printed.
• Optimize with NumPy: The dot product computation is optimized using NumPy's dot() function, and the first 5x5 block of the result is printed.

Python-Numpy Code Editor:

Have another way to solve this solution? Contribute your code (and comments) through Disqus.

Previous: Optimizing element-wise addition of large NumPy arrays.
Next: Optimizing row-wise Mean calculation of large NumPy arrays.