Generalized ufuncs concept in NumPy - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of a simple generalized ufunc call

What is the output of this code using a generalized ufunc with specified core dimensions?

NumPy

import numpy as np

def my_func(x, y):
    return x + y

my_gufunc = np.frompyfunc(my_func, 2, 1)

x = np.array([[1, 2], [3, 4]])
y = np.array([[10, 20], [30, 40]])

result = my_gufunc(x, y)
print(result)

A

[[1 2]
 [3 4]]

B

[[11 12]
 [13 14]]

C

[[10 20]
 [30 40]]

D

[[11 22]
 [33 44]]

💡 Hint

Think about what the function does element-wise on the arrays.

The generalized ufunc applies the function element-wise, adding corresponding elements of x and y.

❓ data_output

intermediate

Shape of output from a generalized ufunc with core dimensions

Given the following generalized ufunc call, what is the shape of the output array?

NumPy

import numpy as np

def func(a, b):
    return a * b

my_gufunc = np.frompyfunc(func, 2, 1)

x = np.ones((3, 4))
y = np.ones((3, 4))

result = my_gufunc(x, y)
print(result.shape)

A (4,)

B (4, 3)

C (3, 4)

D (3,)

💡 Hint

The function is applied element-wise to arrays of the same shape.

The output shape matches the input arrays shape because the function operates element-wise.

🔧 Debug

advanced

Identify the error in this generalized ufunc definition

What error does this code raise when trying to create a generalized ufunc with core dimensions?

NumPy

import numpy as np

def my_func(x, y):
    return x + y

# Incorrect core dimensions specification
my_gufunc = np.frompyfunc(my_func, 2, 1)

x = np.array([1, 2, 3])
y = np.array([4, 5, 6])

result = my_gufunc(x, y)
print(result)

A No error, prints [5 7 9]

B ValueError: operands could not be broadcast together

C SyntaxError: invalid syntax

D TypeError: frompyfunc does not support core dimensions

💡 Hint

Check the capabilities of np.frompyfunc regarding core dimensions.

np.frompyfunc creates simple ufuncs without support for core dimensions, so specifying them causes a TypeError.

🚀 Application

advanced

Using a generalized ufunc to compute matrix multiplication

Which option correctly uses a generalized ufunc to perform matrix multiplication on two 2D arrays?

NumPy

import numpy as np

def matmul(a, b):
    return a @ b

# Assume my_gufunc is defined here

A = np.array([[1, 2], [3, 4]])
B = np.array([[5, 6], [7, 8]])

result = my_gufunc(A, B)
print(result)

A my_gufunc = np.vectorize(matmul, signature='(m,n),(n,p)->(m,p)')

B my_gufunc = np.frompyfunc(matmul, 2, 1)

C my_gufunc = np.frompyfunc(lambda x, y: x * y, 2, 1)

D my_gufunc = np.vectorize(matmul)

💡 Hint

Generalized ufuncs with core dimensions require a signature to specify input/output shapes.

np.vectorize with a signature allows specifying core dimensions for matrix multiplication, unlike frompyfunc.

🧠 Conceptual

expert

Understanding core dimensions in generalized ufuncs

Which statement best describes the role of core dimensions in a generalized ufunc?

A Core dimensions define the output data type of the ufunc.

B Core dimensions specify the fixed dimensions that the ufunc operates on element-wise, while broadcasting over other dimensions.

C Core dimensions are the dimensions that are ignored during the ufunc operation.

D Core dimensions determine the number of input arguments the ufunc accepts.

💡 Hint

Think about how generalized ufuncs handle arrays with different shapes.

Core dimensions are the fixed dimensions the ufunc processes directly; other dimensions are broadcasted over.