How to define and apply a Custom ufunc with Broadcasting in NumPy

NumPy: Universal Functions Exercise-14 with Solution

Custom ufunc with Broadcasting:

Write a Numpy program that defines a custom ufunc that computes 3x + 4y for elements x and y from two arrays, and apply it to 2D arrays.

Sample Solution:

Python Code:

import numpy as np

# Define the custom ufunc
def custom_func(x, y):
    return 3 * x + 4 * y

# Convert the Python function to a NumPy ufunc
custom_ufunc = np.frompyfunc(custom_func, 2, 1)

# Create two 2D NumPy arrays
array_x = np.array([[1, 2], [3, 4]])
array_y = np.array([[5, 6], [7, 8]])

# Apply the custom ufunc to the 2D arrays with broadcasting
result_array = custom_ufunc(array_x, array_y)

# Convert the result to a numeric type array if needed
result_array = result_array.astype(np.float64)

# Print the input arrays and the result array
print("Array X:")
print(array_x)

print("\nArray Y:")
print(array_y)

print("\nResult Array (3x + 4y):")
print(result_array)

Output:

Array X:
[[1 2]
 [3 4]]

Array Y:
[[5 6]
 [7 8]]

Result Array (3x + 4y):
[[23. 30.]
 [37. 44.]]

Explanation:

• Import NumPy Library:
• Import the NumPy library to handle arrays and "ufunc" creation.
• Define the Custom Function:
• Create a custom Python function custom_func that computes 3x + 4y.
• Convert to NumPy ufunc:
• Use np.frompyfunc() to convert the Python function to a NumPy ufunc, specifying that the function takes two input arguments and returns one output.
• Create 2D Arrays:
• Define two 2D NumPy arrays ‘array_x’ and ‘array_y’ with some example data.
• Apply the Custom ufunc:
• Use the custom "ufunc" with broadcasting to compute the result array from 'array_x' and 'array_y'.
• Convert Result to Numeric Type:
• If needed, convert the result array to a specific numeric type (e.g., np.float64) for further numerical operations.
• Finally print the original input arrays and the resulting array to verify the computation.

Python-Numpy Code Editor:

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

Previous: Divide each 2D slice of 3D array by 1D array using np.divide.
Next: Comparing performance of custom ufunc and Python loop in NumPy.