NumPy Interview Questions and Answers for 2025

Element-wise operations are operations that are performed on corresponding elements in two arrays. NumPy provides many functions for performing element-wise operations on arrays. 

Here are some examples of how to perform element-wise operations on NumPy arrays: 

Using NumPy functions: NumPy provides many functions that can be used to perform element-wise operations on arrays. For example, you can use the np.add() function to add two arrays element-wise, the np.subtract() function to subtract one array from another element-wise, and the np.multiply() function to multiply two arrays element-wise. 

import numpy as np 
# Create two arrays 
a = np.array([1, 2, 3]) 
b = np.array([4, 5, 6]) 
# Add the arrays element-wise using the + operator 
c = a + b # element-wise addition: 
print(d) 

#Output: 

#[5, 7, 9] 

# Subtract the arrays element-wise using the - operator 
d = a - b # element-wise subtraction: 
print(d) 

#Output: 

#[-3, -3, -3] 

# Multiply the arrays element-wise using the * operator 
e = a * b # element-wise multiplication:  
print(e) 

#Output: 

#[4, 10, 18] 
# Divide the arrays element-wise using the / operator 
f = a / b # element-wise division:  
print(f) 

#Output: 

#[0.25, 0.4, 0.5] 

# Exponent the arrays element-wise using the 88 operator 
g = a ** b # element-wise exponentiation:  
print(g) 

#Output: 

#[1, 32, 729] 

Using NumPy operators: NumPy also provides many operators that can be used to perform element-wise operations on arrays. For example, you can use the + operator to add two arrays element-wise, the - operator to subtract one array from another element-wise, and the * operator to multiply two arrays element-wise. 

import numpy as np 
# Create two arrays 
a = np.array([1, 2, 3]) 
b = np.array([4, 5, 6]) 
# Add the arrays element-wise using the np.add() function 
h = np.add(a, b): element-wise addition 
print(d) 

#Output: 

#[5, 7, 9] 

# Subtract the arrays element-wise using the np.subtract() function 
i = np.subtract(a, b): element-wise subtraction 
print(d) 

#Output: 

#[-3, -3, -3] 

# Multiply the arrays element-wise using the np.multiply() function 
j = np.multiply(a, b): element-wise multiplication 
print(d) 

#Output: 

#[4, 10, 18] 

# Divide the arrays element-wise using the np.divide() function 
k = np.divide(a, b): element-wise division 
print(d) 

#Output: 

#[0.25, 0.4, 0.5] 

# Exponent the arrays element-wise using the np.power() function 
l = np.power(a, b): element-wise exponentiation 
print(d) 

#Output: 

#[1, 32, 729] 

These functions can be useful when you want to specify additional options, such as the output data type or handling of invalid values (e.g., division by zero). 

You can also use NumPy's universal functions (ufuncs) to perform element-wise operations. Ufuncs are functions that operate element-wise on arrays, like the arithmetic operators and functions described above. Some examples of ufuncs include: 

• np.sin(a): element-wise sine 
• np.cos(a): element-wise cosine 
• np.exp(a): element-wise exponentiation (e raised to the power of each element) 
• np.maximum(a, b): element-wise maximum (returns the larger of the two elements at each position) 

You can find a full list of NumPy's ufuncs in the documentation: https://NumPy.org/doc/stable/reference/ufuncs.html 

To calculate the mean of a NumPy array, you can use the NumPy.mean function. This function takes in the array and returns the mean of the array. 

Here's the basic syntax for NumPy.mean: 

NumPy.mean(a, axis=None, dtype=None, out=None, keepdims=False) 

• a: This is the NumPy array for which you want to calculate the mean. 
• axis: This is an optional argument that specifies the axis along which the mean is calculated. The default value is None, which means that the mean is calculated over all the elements of the array. 
• dtype: This is an optional argument that specifies the data type of the output. The default value is None, which means that the data type is determined from the input array. 
• out: This is an optional argument that specifies an output array in which to store the result. The default value is None. 
• keepdims: This is an optional argument that specifies whether to keep the dimensions of the input array. The default value is False, which means that the dimensions of the input array are flattened. 

Here's an example of how to use NumPy.mean to calculate the mean of a NumPy array: 

import numpy as np 
# Create a NumPy array 
arr = np.array([1, 2, 3, 4, 5]) 
# Calculate the mean of the array 
mean = np.mean(arr) 
print(mean) #

 Output: 3.0 

This will calculate the mean of the array [1, 2, 3, 4, 5] and print it to the console. 

Median 

To calculate the median of a NumPy array, you can use the NumPy.median function. This function takes in the array and returns the median of the array. 

Here's the basic syntax for NumPy.median: 

NumPy.median(a, axis=None, out=None, overwrite_input=False, keepdims=False) 

• a: This is the NumPy array for which you want to calculate the median. 
• axis: This is an optional argument that specifies the axis along which the median is calculated. The default value is None, which means that the median is calculated over all the elements of the array. 
• out: This is an optional argument that specifies an output array in which to store the result. The default value is None. 
• overwrite_input: This is an optional argument that specifies whether to allow the input array to be modified. The default value is False, which means that the input array is not modified. 
• keepdims: This is an optional argument that specifies whether to keep the dimensions of the input array. The default value is False, which means that the dimensions of the input array are flattened. 

Here's an example of how to use NumPy.median to calculate the median of a NumPy array: 

import numpy as np 
# Create a NumPy array 
arr = np.array([1, 2, 3, 4, 5]) 
# Calculate the median of the array 
median = np.median(arr) 
print(median) # 

Output: 3.0 

This will calculate the median of the array [1, 2, 3, 4, 5] and print it to the console. 

To calculate the standard deviation of a NumPy array, you can use the NumPy.std function. This function takes in the array and returns the standard deviation of the array. 

Here's the basic syntax for NumPy.std: 

NumPy.std(a, axis=None, dtype=None, out=None, ddof=0, keepdims=False) 

• a: This is the NumPy array for which you want to calculate the standard deviation. 
• axis: This is an optional argument that specifies the axis along which the standard deviation is calculated. The default value is None, which means that the standard deviation is calculated over all the elements of the array. 
• dtype: This is an optional argument that specifies the data type of the output. The default value is None, which means that the data type is determined from the input array. 
• out: This is an optional argument that specifies an output array in which to store the result. The default value is None. 
• ddof: This is an optional argument that specifies the degrees of freedom to use when calculating the standard deviation. The default value is 0, which means that the standard deviation is calculated using the entire array. 
• keepdims: This is an optional argument that specifies whether to keep the dimensions of the input array. The default value is False, which means that the dimensions of the input array are flattened. 

Here's an example of how to use NumPy.std to calculate the standard deviation of a NumPy array: 

import numpy as np 
# Create a NumPy array 
arr = np.array([1, 2, 3, 4, 5]) 
# Calculate the standard deviation of the array 
std = np.std(arr) 
print(std) # 

Output: 1.4142135623730951 

This will calculate the standard deviation of the array [1, 2, 3, 4, 5] and print it to the console. 

The np.fliplr() function flips an array horizontally (i.e., along the vertical axis), whereas the np.flipud() function flips an array vertically (i.e., along the horizontal axis). 

Here is an example to illustrate the difference between these two functions: 

import numpy as np 
# Create an array 
arr = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) 
print(arr) 

# Output: 

# [[1 2 3] 
# [4 5 6] 
# [7 8 9]] 

# Flip the array horizontally using np.fliplr() 
flipped_arr = np.fliplr(arr) 
print(flipped_arr) 

# Output: 

# [[3 2 1] 
# [6 5 4] 
# [9 8 7]] 

# Flip the array vertically using np.flipud() 
flipped_arr = np.flipud(arr) 
print(flipped_arr) 

# Output: 

# [[7 8 9] 
# [4 5 6] 
# [1 2 3]] 

As you can see, the np.fliplr() function flips the array horizontally, so that the elements on the right side of the array end up on the left side, and the elements on the left side of the array end up on the right side. On the other hand, the np.flipud() function flips the array vertically, so that the elements on the top of the array end up on the bottom, and the elements on the bottom of the array end up on the top. 

I hope this helps to clarify the difference between these two functions! Let me know if you have any questions. 

To create a NumPy array with a sequence of evenly spaced values, you can use the NumPy.linspace function. This function takes in the start value, the end value, and the number of elements, and returns a NumPy array with values evenly spaced between the start and end values. 

Here's the basic syntax for NumPy.linspace: 

NumPy.linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None) 

• start: This is the starting value of the sequence. 
• stop: This is the ending value of the sequence. 
• num: This is the number of elements in the output array. The default value is 50. 
• endpoint: This is a boolean value that specifies whether to include the endpoint in the output array. The default value is True, which means that the endpoint is included. 
• retstep: This is a boolean value that specifies whether to return the step size used to generate the array. The default value is False, which means that the step size is not returned. 
• dtype: This is the data type of the output array. The default value is None, which means that the data type is determined based on the input values. 

Here's an example of how to use NumPy.linspace to create a NumPy array with a sequence of evenly spaced values: 

import numpy as np 
# Create a NumPy array with 10 evenly spaced values from 0 to 1 
arr = np.linspace(0, 1, 10) 
print(arr)  

# Output:  

#[0. 0.11 0.22 0.33 0.44 0.56 0.67 0.78 0.89 1. ] 

This will create a NumPy array with 10 evenly spaced values from 0 to 1, inclusive. 

You can also use the step parameter of the NumPy.arange function to create a NumPy array with evenly spaced values. The NumPy.arange function generates a NumPy array with a range of values, in increments of a given step size. 

Here's the basic syntax for NumPy.arange: 

NumPy.arange(start, stop=None, step=1, dtype=None) 

• start: This is the starting value of the sequence. 
• stop: This is the ending value of the sequence. If not specified, the default value is None, which means that the sequence goes on indefinitely. 
• step: This is the step size between elements. The default value is 1. 
• dtype: This is the data type of the output array. The default value is None, which means that the data type is determined based on the input values. 

Here's an example of how to use NumPy.arange to create a NumPy array with a sequence of evenly spaced values: 

import numpy as np 
# Create a NumPy array with 10 evenly spaced values from 0 to 1, in increments of 0.1 
arr = np.arange(0, 1.1, 0.1) 
print(arr)  

# Output:  

#[0. 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1. ] 

To create a NumPy array with a sequence of logarithmically spaced values, you can use the NumPy.logspace function. This function takes in the start value, the end value, and the number of elements, and returns a NumPy array with logarithmically spaced values between the start and end values. 

Here's the basic syntax for NumPy.logspace: 

NumPy.logspace(start, stop, num=50, endpoint=True, base=10.0, dtype=None) 

• start: This is the starting value of the sequence, expressed as the logarithm of the value. 
• stop: This is the ending value of the sequence, expressed as the logarithm of the value. 
• num: This is the number of elements in the output array. The default value is 50. 
• endpoint: This is a boolean value that specifies whether to include the endpoint in the output array. The default value is True, which means that the endpoint is included. 
• base: This is the base of the logarithm used to generate the array. The default value is 10. 
• dtype: This is the data type of the output array. The default value is None, which means that the data type is determined based on the input values. 

import numpy as np 
# Create a NumPy array with 10 logarithmically spaced values from 1 to 100 
arr = np.logspace(0, 2, 10) 
print(arr) 

# Output: 

[ 1. 1.66810054 2.7825594 4.64158883 7.74263683 
# 12.91549665 21.5443469 36.01778261 59.94842503 100. ] 

This will create a NumPy array with 10 logarithmically spaced values from 1 to 100, inclusive. The values are spaced such that each value is the base-10 logarithm of the value. 

You can specify a different base for the logarithm by using the base parameter: 

import numpy as np 
# Create a NumPy array with 10 logarithmically spaced values from 1 to 100, using base 2 
arr = np.logspace(0, 2, 10, base=2) 
print(arr) 

# Output: 

[1. 1.18920712 1.41421356 1.68179283 2. 2.37841423 
# 2.82842712 3.36358566 4. 4.75682846] 

This will create a NumPy array with 10 logarithmically spaced values from 1 to 100, using base 2. The values are spaced such that each value is the base-2 logarithm of the value. 

You can also use the NumPy.geomspace function to create a NumPy array with logarithmically spaced values. The NumPy.geomspace function generates a NumPy array with a sequence of logarithmically spaced values between a start value and an end value, in increments of a geometric series. 

Here's the basic syntax for NumPy.geomspace: 

NumPy.geomspace(start, stop, num=50, endpoint=True, dtype=None) 

• start: This is the starting value of the sequence. 
• stop: This is the ending value of the sequence. 
• num: This is the number of elements in the output array. The default value is 50. 
• endpoint: This is a boolean value that specifies whether to include the endpoint in the output array. The default value is True, which means that the endpoint is included. 
• dtype: This is the data type of the output array. The default value is None, which means that the data type is determined based on the input values. 

Here's an example of how to use NumPy.geomspace to create a NumPy array with a sequence of logarithmically spaced values: 

import numpy as np 
# Create a NumPy array with 10 logarithmically spaced values from 1 to 100 
arr = np.geomspace(1, 100, 10) 
print(arr) 

# Output:

[ 1. 3.16227766 10. 31.6227766 100. ] 

This will create a NumPy array with 5 logarithmically spaced values from 1 to 100, in increments of a geometric series. 

This is a common yet one of the most important NumPy interview questions and answers for experienced professionals, don't miss this one.

To create a NumPy array with random values, you can use the NumPy.random module. The NumPy.random module contains a number of functions for generating random numbers, and you can use these functions to create a NumPy array with random values 

Here are some examples of common functions you might use: 

random: The random function generates random floats between 0 and 1. For example: 

import numpy as np 
# Create a 3x3 array with random values between 0 and 1 
random_array = np.random.random((3, 3)) 
print(random_array) 

This would output a 3x3 array with random values between 0 and 1: 

[[0.1234 0.5678 0.9101] 
 [0.2345 0.6789 0.1234] 
 [0.3456 0.7890 0.2345]] 

randint: Generates random integers within a given range. For example, np.random.randint(0, 10, (3, 3)) would generate a 3x3 array of random integers between 0 and 9. 

import numpy as np 
# Create a 3x3 array with random integers between 0 and 9 
random_array = np.random.randint(0, 10, (3, 3)) 
print(random_array) 

This would output a 3x3 array with random integers between 0 and 9: 

[[4 7 2] 
 [9 3 5] 
 [6 2 8]] 

normal: Generates random values that are normally distributed (i.e., with a bell curve shape). You can specify the mean and standard deviation of the distribution. For example, np.random.normal(0, 1, (3, 3)) would generate a 3x3 array of random values with a mean of 0 and a standard deviation of 1. 

import numpy as np 
# Create a 3x3 array with random values that are normally distributed 
# with a mean of 0 and a standard deviation of 1 
random_array = np.random.normal(0, 1, (3, 3)) 
print(random_array) 

This would output a 3x3 array with random values that are normally distributed with a mean of 0 and a standard deviation of 1: 

[[-0.5678 0.2345 0.9101] 
 [ 0.6789 -1.1234 0.1234] 
 [ 0.3456 0.7890 -0.2345]] 

choice: Generates random values from a given sequence (e.g., a list or array). For example, np.random.choice([0, 1, 2, 3], (3, 3)) would generate a 3x3 array of random values, with each value being chosen from the sequence [0, 1, 2, 3]. 

import numpy as np 
# Create a 3x3 array with random values chosen from the sequence [0, 1, 2, 3] 
random_array = np.random.choice([0, 1, 2, 3], (3, 3)) 
print(random_array) 

This would output a 3x3 array with random values chosen from the sequence [0, 1, 2, 3]: 

[[2 1 3] 
 [3 1 0] 
 [0 2 1]] 

These are just a few examples of the functions available in the NumPy.random module. There are many other functions available in the NumPy.random module for generating different types of random numbers. For example, you can use the randint function to create an array of random integers, or the normal function to create an array of random values that are normally distributed. You can find more information about these functions in the NumPy documentation.

The np.where function is a way to perform element-wise operations on NumPy arrays based on a condition. It takes three arguments: 

A condition: This can be either a single boolean value, or a boolean array of the same shape as the arrays you want to operate on. This is used to determine which elements should be operated on. For example, if you want to set all negative values in an array to zero, the condition could be a < 0, which would return a boolean array of the same shape as a, with True for negative elements and False for non-negative elements. 

An array or a scalar value to use if the condition is True: This is the value that will be used for elements where the condition is True. If you pass an array, it should have the same shape as the arrays you want to operate on. If you pass a scalar value, it will be used for all elements where the condition is True. 

An array or a scalar value to use if the condition is False: This is the value that will be used for elements where the condition is False. If you pass an array, it should have the same shape as the arrays you want to operate on. If you pass a scalar value, it will be used for all elements where the condition is False. 

Here's an example of how you can use np.where to set all negative values in an array to zero: 

import numpy as np 
# Initialize an array with some negative values 
a = np.array([-1, 4, -9, 2, -5, 8]) 
# Use np.where to set all negative values to zero 
result = np.where(a < 0, 0, a) 
print(result) # [0 4 0 2 0 8] 

In this example, the condition is a < 0, which returns a boolean array [True, False, True, False, True, False]. The np.where function then uses this boolean array to select the elements of a where the condition is True (i.e., the negative elements) and sets them to zero. The elements where the condition is False (i.e., the non-negative elements) are left unchanged. 

You can also use np.where to perform operations on multiple arrays. For example, here's how you can add two arrays element-wise, but only add the corresponding elements if both are positive: 

import numpy as np 
# Initialize two arrays 
a = np.array([-1, 4, -9, 2, -5, 8]) 
b = np.array([3, -4, 7, -2, 5, -8]) 
# Use np.where to add the arrays element-wise, but only add the elements if both are positive 
result = np.where((a > 0) & (b > 0), a + b, 0) 
print(result) # [0 8 0 4 0 16] 

In this example, the np.where function uses the boolean array [False, True, False, True, False, True] to select the elements of a and b where the condition is True (i.e., the positive elements). It then adds these elements element-wise and returns a new array with the results. The elements where the condition is False (i.e., the non-positive elements) are set to zero. 

You can use any condition you like in the np.where function, as long as it returns a boolean array or a single boolean value. You can also use the np.where function to perform any element-wise operation, not just setting values to a specific array or scalar. 

For example, you could use the np.where function to multiply two arrays element-wise, but only multiply the corresponding elements if both are even: 

import numpy as np 
# Initialize two arrays 
a = np.array([2, 4, 6, 8, 10, 12]) 
b = np.array([1, 2, 3, 4, 5, 6]) 
# Use np.where to multiply the arrays element-wise, but only multiply the elements if both are even 
result = np.where((a % 2 == 0) & (b % 2 == 0), a * b, 0) 
print(result) #[0 8 0 32 0 72] 

NumPy is a powerful library for working with numerical data in Python. It provides a number of functions and tools for working with arrays, which are N-dimensional grid-like data structures. NumPy arrays are particularly useful for performing mathematical and statistical operations, as they allow you to perform element-wise operations and operate on entire arrays rather than individual elements. 

One of the data types that can be stored in a NumPy array is the object dtype. This dtype is used to store elements that are of a more general Python object type, rather than a specific numerical type such as float or int. When an array has a dtype object, it can store elements of any Python object type, including strings. 

To perform string operations on a NumPy array of dtype objects, you can use NumPy's string functions, which are available in the NumPy.char module. These functions allow you to perform a variety of operations on strings, such as converting them to uppercase or lowercase, capitalizing the first letter, stripping leading or trailing whitespace, splitting strings on a delimiter, and joining strings with a separator. 

Here's an example of using some of these string functions on a NumPy array of dtype object: 

import numpy as np 
# Create a NumPy array with dtype object 
arr = np.array([' cat ', 'DOG', 'birD', 'Fish '], dtype=object) 
# Convert all strings to lowercase 
arr_lower = np.char.lower(arr) 
# Capitalize the first letter of each string 
arr_capitalized = np.char.capitalize(arr_lower) 
# Strip leading and trailing whitespace 
arr_stripped = np.char.strip(arr_capitalized) 
# Split strings on space character 
arr_split = np.char.split(arr_stripped, sep=' ') 
# Join strings with '-' character 
arr_joined = np.char.join('-', arr_split) 
print(arr_joined) # Output: ['cat' 'dog' 'bird' 'fish'] 
Keep in mind that NumPy's string functions operate element-wise on the array, meaning that they are applied to each element in the array separately. This allows you to perform the same operation on all the elements in the array with a single function call. 

NumPy's object dtype is used to store elements that are of a more general Python object type, rather than a specific numerical type (such as float or int). When an array has a dtype object, it can store elements of any Python object type, including strings. 

To perform string operations on a NumPy array of dtype object, you can use NumPy's string functions, which are available in the NumPy.char module. These functions include upper, lower, capitalize, strip, split, join, and many others. Here's an example of using the upper function to convert all the strings in a NumPy array to uppercase: 

import numpy as np 
# Create a NumPy array with dtype object 
arr = np.array(['cat', 'dog', 'bird', 'fish'], dtype=object) 
# Convert all strings to uppercase 
arr_upper = np.char.upper(arr) 
print(arr_upper) # Output: ['CAT' 'DOG' 'BIRD' 'FISH'] 

Keep in mind that NumPy's string functions operate element-wise on the array, meaning that they are applied to each element in the array separately. 

Don't be surprised if this question pops up as one of the top NumPy programming interview questions for data science in your next interview.

Missing or invalid data, also known as "missing values," can occur in a dataset for a variety of reasons. For example, a measurement might be missing because it was not taken, or a value might be invalid because it falls outside the acceptable range for that variable. When working with numerical data, it is important to identify and handle missing values appropriately to ensure that they do not bias your analysis or lead to errors. 

In NumPy, there are a few different ways to identify and handle missing values. One approach is to use the NumPy.isnan function, which returns a Boolean array indicating which elements in an array are NaN (Not a Number). You can use this function to identify missing values and then replace them with a suitable substitute value, such as the mean or median of the data. This special floating-point value is used to represent missing or undefined numeric data, and it is not considered equal to any other value (including itself). You can use the NumPy.isnan function to identify elements in an array that have the value NaN and then replace them with a suitable substitute value. 

Here's an example of using NumPy.isnan to identify and replace missing values in a NumPy array: 

import numpy as np 
# Create a NumPy array with some missing values 
arr = np.array([1, 2, 3, np.nan, 5, 6, 7, 8]) 
# Identify missing values with NumPy.isnan 
mask = np.isnan(arr) 
# Replace missing values with the mean of the data 
mean = arr[~mask].mean() 
arr[mask] = mean 
print(arr) # Output: [1. 2. 3. 4.5 5. 6. 7. 8.] 

Another approach to handling missing values in NumPy is to use the NumPy.ma module, which provides tools for working with masked arrays. A masked array is an array with a separate Boolean mask that indicates which elements are missing or invalid. You can use the NumPy.ma.masked_invalid function to create a masked array from an existing array, and then use the mask to perform operations on the data while ignoring the missing values. 

Here's an example of using a masked array to perform statistical operations while ignoring missing values: 

import numpy as np 
# Create a NumPy array with some missing values 
arr = np.array([1, 2, 3, np.nan, 5, 6, 7, 8]) 
# Create a masked array from the data 
masked_arr = np.ma.masked_invalid(arr) 
# Calculate the mean of the data, ignoring missing values 
mean = masked_arr.mean() 
print(mean) # Output: 4.5 

In this example, the NumPy.ma.masked_invalid function is used to create a masked array from the original array, with a mask that indicates which elements are NaN. The mean method of the masked array is then used to calculate the mean of the data, ignoring the missing values. 

There are many other ways to handle missing values in NumPy, and which approach you choose will depend on the specifics of your data and the goals of your analysis. 

NumPy provides a number of functions for reading and writing arrays to and from file, allowing you to easily save and load data in a variety of formats. Some of the most commonly used functions for file I/O (input/output) with NumPy arrays include: 

NumPy.save: Saves a single NumPy array to a binary file with .npy extension. The NumPy.save function takes two arguments: the filename, and the array to be saved. It saves the array to a file with the specified name and a .npy extension. The file is a binary file that contains the data and metadata of the array, including its shape, data type, and other attributes. 

import numpy as np 
# Create a NumPy array 
arr = np.array([1, 2, 3, 4, 5]) 
# Save the array to a .npy file 
np.save('arr.npy', arr) 

NumPy.savez: Saves multiple NumPy arrays to a single .npz file, which is a ZIP archive containing the arrays. The NumPy.savez function takes a filename and a sequence of arrays to be saved, and it stores the arrays in a ZIP archive with the specified name and a .npz extension. The arrays are stored in the archive with their names as keys, allowing you to retrieve them by key when you load the file. 

import numpy as np 
# Create two NumPy arrays 
arr1 = np.array([1, 2, 3]) 
arr2 = np.array([4, 5, 6]) 
# Save the arrays to a .npz file 
np.savez('arrays.npz', arr1=arr1, arr2=arr2) 

NumPy.savetxt: Saves a NumPy array to a text file, with the option to specify the delimiter and precision. The NumPy.savetxt function takes a filename, the array to be saved, and a number of optional arguments for formatting the data. It saves the array to a text file with the specified name, using the specified delimiter to separate the values and the specified precision to control the number of decimal places. 

import numpy as np 
# Create a NumPy array 
arr = np.array([[1, 2, 3], [4, 5, 6]]) 
# Save the array to a text file with space-separated values 
np.savetxt('arr.txt', arr, delimiter=' ') 

NumPy.load: Loads a single NumPy array from a .npy file. The NumPy.load function takes a single argument, the filename of the .npy file to be loaded, and returns the array stored in the file. It automatically reconstructs the array from the data and metadata in the file, including its shape, data type, and other attributes. 

import numpy as np 
# Load the array from a .npy file 
loaded_arr = np.load('arr.npy') 
print(loaded_arr) # Output: [1 2 3 4 5] 

Using these functions, you can easily save and load NumPy arrays to and from a variety of file formats, including binary files, text files, and ZIP archives. This can be useful for storing data for later use, sharing data with others, or for reading in data from external sources. 

One of the most frequently posed NumPy scenario based interview questions, be ready for this conceptual question.

To compute the moving average of an array in NumPy, you can use the NumPy.convolve function with the 'valid' mode. 

Here is an example of how you can use NumPy.convolve to compute the moving average of an array with a window size of 3: 

import numpy as np 
def moving_average(arr, window_size): 
return np.convolve(arr, np.ones(window_size)/window_size, mode='valid') 
arr = np.array([1, 2, 3, 4, 5, 6, 7, 8, 9]) 
moving_average(arr, 3) 

This will return the moving average of the array with a window size of 3, which is: [2. 3. 4. 5. 6. 7.] 

The NumPy.convolve function computes the discrete, linear convolution of two one-dimensional sequences. In this case, we are using it to compute the moving average of an array by treating the array as the input sequence and a window of size window_size as the second sequence. 

The mode parameter specifies the size and shape of the output, and we are using the 'valid' mode, which means that the output will only contain parts of the convolution that are computed without the zero-padding. This results in an output that is (len(arr) - window_size + 1) elements long. 

The np.ones(window_size)/window_size is used as the second sequence to compute the moving average. It is a window of size window_size filled with ones and divided by window_size to normalize the output. 

For example, if arr is [1, 2, 3, 4, 5, 6, 7, 8, 9] and window_size is 3, the convolution will be computed as follows: 

(1*1 + 2*1 + 3*1)/3 = 2 
(2*1 + 3*1 + 4*1)/3 = 3 
(3*1 + 4*1 + 5*1)/3 = 4 
(4*1 + 5*1 + 6*1)/3 = 5 
(5*1 + 6*1 + 7*1)/3 = 6 
(6*1 + 7*1 + 8*1)/3 = 7 
And the result will be [2, 3, 4, 5, 6, 7]. 

The in1d function in NumPy is used to test whether each element of one array is contained in another array. It takes as input two arrays, and returns a boolean array with the same shape as the first array, indicating whether each element is contained in the second array. 

For example, consider the following code: 

import numpy as np 
# Create two arrays 
a = np.array([1, 2, 3, 4]) 
b = np.array([2, 4, 6, 8]) 
# Test whether each element of a is contained in b 
result = np.in1d(a, b) 
print(result) # prints [False True False True] 

In this example, the in1d function tests whether each element of the array a is contained in the array b. The returned boolean array, result, has the same shape as a and indicates whether each element is contained in b. 

You can use the in1d function to find the common elements between two arrays, or to filter one array based on the values in another array. For example: 

import numpy as np 
# Create two arrays 
a = np.array([1, 2, 3, 4]) 
b = np.array([2, 4, 6, 8]) 
# Find the common elements between a and b 
common_elements = a[np.in1d(a, b)] 
print(common_elements) # prints [2 4] 
# Filter a based on the values in b 
filtered_a = a[np.in1d(a, b, invert=True)] 
print(filtered_a) # prints [1 3] 

The in1d function in NumPy is used to test whether each element of one array is contained in another array. It is useful for a variety of tasks, such as finding the common elements between two arrays, filtering one array based on the values in another array, and performing set operations on arrays. 

For example, you can use the in1d function to find the common elements between two arrays: 

import numpy as np 
# Create two arrays 
a = np.array([1, 2, 3, 4]) 
b = np.array([2, 4, 6, 8]) 
# Find the common elements between a and b 
common_elements = a[np.in1d(a, b)] 
print(common_elements) # prints [2 4] 

You can also use the in1d function to filter one array based on the values in another array: 

import numpy as np 
# Create two arrays 
a = np.array([1, 2, 3, 4]) 
b = np.array([2, 4, 6, 8]) 
# Filter a based on the values in b 
filtered_a = a[np.in1d(a, b, invert=True)] 
print(filtered_a) # prints [1 3] 

The invert keyword argument specifies whether to invert the test (i.e., whether to return elements that are not contained in the second array). 

You can also use the in1d function to perform set operations on arrays, such as finding the elements that are present in one array but not the other: 

import numpy as np 
# Create two arrays 
a = np.array([1, 2, 3, 4]) 
b = np.array([2, 4, 6, 8]) 

# Find the elements that are present in a but not b 

difference = a[np.in1d(a, b, invert=True)] 
print(difference) # prints [1 3] 
# Find the elements that are present in b but not a 
difference = b[np.in1d(b, a, invert=True)] 
print(difference) # prints [6 8] 

To compute the rank of a matrix in NumPy, you can use the linalg.matrix_rank function from the NumPy.linalg module. This function takes a matrix as input and returns its rank, which is defined as the number of linearly independent rows or columns in the matrix. 

Here is an example of how to use this function: 

import numpy as np 
# Create a matrix 
matrix = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) 
# Compute the rank of the matrix 
rank = np.linalg.matrix_rank(matrix) 
# Print the rank of the matrix 
print(rank) # Output: 2 

Note that the rank of a matrix is generally less than or equal to its number of rows and columns. A matrix with full rank is said to be non-singular, while a matrix with rank less than its number of rows and columns is said to be singular. 

You can also use the linalg.matrix_rank function to compute the rank of a multi-dimensional array, by specifying the axis parameter, which indicates the axis along which the rank is to be computed. For example: 

import numpy as np 
# Create a 3D array 
array = np.array([[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [10, 11, 12]]]) 
# Compute the rank of the array along axis 0 (depth) 
rank = np.linalg.matrix_rank(array, axis=0) 
# Print the rank of the array 
print(rank) # Output: [[2, 2, 2], [2, 2, 2]] 

By default, the linalg.matrix_rank function uses a singular value decomposition (SVD) to compute the rank of the matrix. You can also specify a different algorithm using the method parameter, such as 'svd', 'qr', or 'cholesky'.

One of the most frequently posed NumPy scenario based interview questions, be ready for this conceptual question.

Here's how you can perform linear algebra operations on NumPy arrays using NumPy's built-in functions: 

To calculate the dot product of two NumPy arrays, you can use the NumPy.dot function. This function takes in the two arrays and returns the dot product of the arrays. 

Here's the basic syntax for NumPy.dot: 

NumPy.dot(a, b, out=None) 

• a: This is the first NumPy array. 
• b: This is the second NumPy array. 
• out: This is an optional argument that specifies an output array in which to store the result. The default value is None. 

Here's an example of how to use NumPy.dot to calculate the dot product of two NumPy arrays: 

import numpy as np 
# Create two NumPy arrays 
a = np.array([1, 2, 3]) 
b = np.array([4, 5, 6]) 
# Calculate the dot product of the arrays 
dot_product = np.dot(a, b) 
print(dot_product) # Output: 32 

This will calculate the dot product of the arrays [1, 2, 3] and [4, 5, 6] and print it to the console. 

Matrix Multiplication 

To perform matrix multiplication on two NumPy arrays, you can use the NumPy.matmul function. This function takes in the two arrays and returns the result of the matrix multiplication. 

Here's the basic syntax for NumPy.matmul: 

NumPy.matmul(a, b, out=None) 

• a: This is the first NumPy array. 
• b: This is the second NumPy array. 
• out: This is an optional argument that specifies an output array in which to store the result. The default value is None. 

Here's an example of how to use NumPy.matmul to perform matrix multiplication on two NumPy arrays: 

import numpy as np 
# Create two NumPy arrays 
a = np.array([[1, 2], [3, 4]]) 
b = np.array([[5, 6], [7, 8]]) 
# Perform matrix multiplication on the arrays 
matrix_multiplication = np.matmul(a, b) 
print(matrix_multiplication) # Output: [[19 22] [43 50]] 

This will perform matrix multiplication on the arrays [[1, 2], [3, 4]] and [[5, 6], [7, 8]] and print the result to the console. 

Singular Value Decomposition 

To perform singular value decomposition on a NumPy array, you can use the NumPy.linalg.svd function. This function takes in the array and returns the singular value decomposition of the array. 

are 

Here's the basic syntax for NumPy.linalg.svd: 

NumPy.linalg.svd(a, full_matrices=True, compute_uv=True, hermitian=False) 

• a: This is the NumPy array for which you want to perform singular value decomposition. 
• full_matrices: This is an optional argument that specifies whether to return full-sized matrices. The default value is True, which means that full-sized matrices are returned. 
• compute_uv: This is an optional argument that specifies whether to compute the matrices U and V. The default value is True, which means that the matrices are computed. 
• hermitian: This is an optional argument that specifies whether to perform singular value decomposition on a Hermitian matrix. The default value is False, which means that singular value decomposition is not performed on a Hermitian matrix. 

import numpy as np 
# Create a NumPy array 
a = np.array([[1, 2], [3, 4]]) 
# Perform singular value decomposition on the array 
U, S, V = np.linalg.svd(a) 
print(U) # Output: [[-0.40455358 -0.9145143 ] [-0.9145143 0.40455358]] 
print(S) # Output: [5.4649857 0.36596619] 
print(V) # Output: [[-0.57604844 -0.81741556] [ 0.81741556 -0.57604844]] 

This will perform singular value decomposition on the array [[1, 2], [3, 4]] and print the matrices U, S, and V to the console. The matrix U is the left singular matrix, S is the singular values, and V is the right singular matrix. 

Here is a more detailed explanation of masked arrays in NumPy:

Creating masked arrays: There are several ways to create a masked array in NumPy. The most basic way is to use the np.ma.masked_array function, which takes a NumPy array and a mask as inputs, and returns a masked array with the same data as the input array, but with masked values indicated by the mask. The mask is a Boolean array with the same shape as the input array, where True indicates a masked value and False indicates a valid value.

For example:

import numpy as np 
# Create a NumPy array with some invalid data 
data = np.array([1, 2, -999, 4, 5]) 
# Create a mask to identify the invalid data 
mask = np.array([False, False, True, False, False]) 
# Create a masked array from the data and mask 
masked_array = np.ma.masked_array(data, mask) 
print(masked_array) # Output: [1 2 -- 4 5] 

In the above example, the third element of the input array (-999) is marked as invalid using the mask, and is represented as "--" in the masked array.

Alternatively, you can use the np.ma.masked_where function to create a masked array by specifying a condition that determines which values in the input array should be masked. For example:

import numpy as np 
# Create a NumPy array with some invalid data 
data = np.array([1, 2, -999, 4, 5]) 
# Create a masked array where the invalid data is masked 
masked_array = np.ma.masked_where(data < 0, data) 
print(masked_array) # Output: [1 2 -- 4 5] 

In this example, the masked array is created by masking all values in the input array that are less than 0.

Accessing and manipulating masked arrays: Once you have created a masked array, you can access and manipulate its data using various functions and methods provided by NumPy's masked array module (np.ma). For example, you can use the .mask attribute to access the mask of a masked array, or the .data attribute to access the underlying data.

You can also use various functions and methods to perform operations on masked arrays. For example, you can use the np.ma.mean function to compute the mean of a masked array, which will automatically exclude the masked values from the calculation. You can also use the .filled method to fill the masked values with a specified value, or the .compressed method to return a flattened version of the array with the masked values removed.

Here is an example that demonstrates some of these operations:

import numpy as np 
# Create a NumPy array with some invalid data 
data = np.array([1, 2, -999, 4, 5]) 
# Create a masked array where the invalid data is masked 
masked_array = np.ma.masked_where(data < 0, data) 
# Access the mask of the masked array 
print(masked_array.mask) # Output: [False, False, True, False, False] 
# Access the underlying data of the masked array 
print(masked_array.data) # Output: [1, 2, -999, 4, 5] 
# Compute the mean of the masked array (excludes the masked value) 
print(np.ma.mean(masked_array)) # Output: 3.0 
# Fill the masked values with 0 
filled_array = masked_array.filled(0) 
print(filled_array) # Output: [1, 2, 0, 4, 5] 
# Remove the masked values 
compressed_array = masked_array.compressed() 
print(compressed_array) # Output: [1, 2, 4, 5] 

In this example, we first create a masked array using the np.ma.masked_where function, and then access the mask and underlying data using the .mask and .data attributes, respectively. We then use the np.ma.mean function to compute the mean of the masked array, which excludes the masked value (-999) from the calculation. We then use the .filled method to fill the masked values with 0, and the .compressed method to remove the masked values from the array.

The np.memmap function allows you to create a NumPy array that is stored in a file on disk, rather than in memory. This can be useful if you have a large array that does not fit in memory, but you still want to perform operations on it. 

When you create a memory-mapped array using np.memmap, you specify the following arguments: 

• filename: The name of the file on disk where the array will be stored. This file will be created if it does not already exist. 
• dtype: The data type of the array. This can be any valid NumPy data type, such as float64, int32, bool, etc. 
• mode: The mode in which the file should be opened. This can be either 'r' (read-only), 'r+' (read and write), or 'w+' (write and read). 
• shape: The shape of the array. This should be a tuple of integers, indicating the size of the array along each dimension. 
• offset: The offset (in bytes) from the start of the file where the array data should be stored. This is optional and defaults to 0. 

For example, to create a memory-mapped array with shape (3, 3) and dtype float64, stored in the file my_array.dat, you could use the following code: 

import numpy as np 
# Create a memory-mapped array with shape (3, 3) and dtype float64, 
# stored in the file 'my_array.dat' 
array = np.memmap('my_array.dat', dtype='float64', mode='w+', shape=(3, 3)) 

This will create a memory-mapped array with shape (3, 3) and dtype float64, stored in the file my_array.dat. The array will be created with all elements initialized to 0. 

To set the values of the array, you can use an assignment like you would with any other NumPy array: 

# Set the values of the array. 
array[:] = np.random.random((3, 3)) 

This will set the values of the array to random values between 0 and 1. 

It's important to note that the changes you make to a memory-mapped array are not immediately persisted to disk. To ensure that the changes are written to disk, you can use the flush method. 

# Flush the changes to the disk 
array.flush() 

Once you have finished making changes to the array, you can close the file by deleting the array. 

# Close the file 
del array 

To reopen the memory-mapped array, you can use the np.memmap function again, specifying the same filename, dtype, and shape arguments, and setting the mode to "r" (read-only) or "r+" (read and write): 

# Re-open the array in read-only mode 
array = np.memmap('my_array.dat', dtype='float64', mode='r', shape=(3, 3)) 
# Print the values of the array 
print(array) 

This will reopen the memory-mapped array

The np.linalg module is a submodule of NumPy that provides functions for performing advanced linear algebra operations on NumPy arrays. Some examples of functions you might use from np.linalg include: 

np.linalg.inv: computes the inverse of a square matrix. The inverse of a matrix A is a matrix A_inv such that A_inv * A = I, where I is the identity matrix. The inverse of a matrix is only defined for square matrices. 

import numpy as np 
# Create a square matrix 
A = np.array([[1, 2], 
[3, 4]]) 
# Compute the inverse of the matrix 
A_inv = np.linalg.inv(A) 
print(A_inv) 

Output: 

[[-2. 1. ][ 1.5 -0.5]] 

np.linalg.svd: Computes the singular value decomposition (SVD) of a matrix. The SVD of a matrix A is a factorization of the form A = U * S * V^T, where U and V are orthogonal matrices and S is a diagonal matrix. The SVD is a powerful tool for analyzing the structure of a matrix, and is often used in machine learning and data analysis. 

import numpy as np 
# Create a matrix 
A = np.array([[1, 2, 3], 
[4, 5, 6]]) 
# Compute the SVD of the matrix 
U, S, V_T = np.linalg.svd(A) 
print(f"U: {U}") 
print(f"S: {S}") 
print(f"V^T: {V_T}") 

Output: 

U: [[-0.3863177 -0.92236578] [-0.92236578 0.3863177 ]] 
S: [9.508032 0.77286964] 
V^T: [[-0.42866713 -0.56630692 -0.7039467 ] 
 [ 0.80596391 0.11238241 -0.58119908] 
 [ 0.40824829 -0.81649658 0.40824829]] 

np.linalg.eig: Computes the eigenvalues and eigenvectors of a square matrix. The eigenvalues and eigenvectors of a matrix A are values and vectors such that A * v = lambda * v, where lambda is an eigenvalue and v is an eigenvector. The eigenvalues and eigenvectors of a matrix are often used to analyze its properties and behavior. 

import numpy as np 
# Create a square matrix 
A = np.array([[1, 2], 
[3, 4]]) 
# Compute the eigenvalues and eigenvectors of the matrix 
eigenvalues, eigenvectors = np.linalg.eig(A) 
print(f"Eigenvalues: {eigenvalues}") 
print(f"Eigenvectors: {eigenvectors}") 

Output: 

Eigenvalues: [-0.37228132 5.37228132] 
Eigenvectors: [[-0.82456484 -0.41597356] 
 [ 0.56576746 -0.90937671]] 

np.linalg.lstsq is a function that solves a linear least-squares problem. Given a matrix A and a vector b, it computes the vector x that minimizes the residual ||A * x - b||_2, where ||x||_2 is the Euclidean norm of x. This is often used to fit a linear model to data. 

import numpy as np 
# Generate some synthetic data 
x = np.linspace(0, 1, 10) 
y = 2 * x + 1 + np.random.normal(0, 0.1, 10) 
# Fit a linear model to the data 
A = np.vstack((x, np.ones(len(x)))).T 
m, c = np.linalg.lstsq(A, y, rcond=None)[0] 
print(f"Slope: {m}") 
print(f"Intercept: {c}") 

Output: 

Slope: 2.000390069852736 
Intercept: 0.9991791312402291 

np.linalg.norm: Computes the norm of a matrix or vector. The norm of a matrix or vector is a measure of its size or length. There are several different types of norms that can be computed, including the Euclidean norm, the Frobenius norm, and the max norm. 

import numpy as np 
# create a matrix 
A = np.array([[1, 2], 
[3, 4]]) 
# Compute the Frobenius norm of the matrix 
frobenius_norm = np.linalg.norm(A, 'fro') 
print(f"Frobenius norm: {frobenius_norm}") 

Output: 

Frobenius norm: 5.477225575051661 

np.linalg.solve: Solves a linear system of equations. Given a matrix A and a vector b, this function computes the vector x such that A * x = b. This is often used to solve systems of linear equations, such as those that arise in linear regression or least-squares fitting. 

import numpy as np 
# Create a matrix and a vector 
A = np.array([[1, 2], 
[3, 4]]) 
b = np.array([5, 6]) 
# Solve the linear system A * x = b 
x = np.linalg.solve(A, b) 
print(x) 

Output: 

[-4. 4.5] 

The tofile method of a NumPy array writes the binary representation of the array to a file. The binary representation of an array is the sequence of bytes that represents the elements of the array in memory. The tofile method writes these bytes to a file so that the array can be reconstructed later by reading the bytes back from the file. 

The tofile method has the following syntax: 

array.tofile(file, sep="", format="%s") 

Here is what each of the arguments does: 

• file: This is the file object to which the data is written. It must be opened in binary mode (using 'wb' for writing or 'rb' for reading). 
• sep: This is an optional separator string that is written between elements. If specified, it is inserted as a string between each element when the data is written to the file. 
• format: This is a format string that specifies how the data should be written. The default is to write the data as ASCII characters, but you can specify a different format if you prefer. For example, you can use '%d' to write the data as integers, or '%f' to write the data as floating-point numbers. 

Here is an example of how to use tofile to write a NumPy array to a binary file: 

import numpy as np 
# Create a NumPy array 
data = np.array([1, 2, 3, 4, 5], dtype=np.int32) 
# Open a binary file for writing 
with open("data.bin", "wb") as f: 
# Write the array to the file 
data.tofile(f) 

This will write the binary representation of the array to the file data.bin. 

fromfile 

The fromfile function reads a NumPy array from a binary file. It reads the binary representation of the array from the file and then reconstructs the array by interpreting the bytes as elements of the specified data type. 

The fromfile function has the following syntax: 

np.fromfile(file, dtype=float, count=-1, sep='') 

Here is what each of the arguments does: 

• file: This is the file object from which the data is read. It must be opened in binary mode (using 'wb' for writing or 'rb' for reading). 
• dtype: This is the data type of the elements in the array. The data type determines how the bytes in the file are interpreted as elements of the array. For example, if you specify dtype=np.int32, the bytes will be interpreted as 32-bit integers. 
• count: This is the number of elements to read from the file. If count is negative, all remaining elements in the file are read. 
• sep: This is an optional separator string that is used to skip over elements in the file. If specified, the function will skip over any element in the file that is followed by the separator string. 

Here is an example of how to use fromfile to read a NumPy array from a binary file: 

import numpy as np 
# Open a binary file for reading 
with open("data.bin", "rb") as f: 
# Read the array from the file 
data = np.fromfile(f, dtype=np.int32) 
print(data) # prints [1 2 3 4 5] 

This will read the binary representation of the array from the file data.bin, and then interpret the bytes as 32-bit integers to reconstruct the array. The resulting array will be printed to the console. 

Keep in mind that tofile and fromfile are low-level functions and are not typically used in practice. Instead, it is more common to use NumPy's save and load functions, which allow you to save and load NumPy arrays to and from files in a more flexible and convenient way. 

A staple in NumPy advanced interview questions and answers, be prepared to answer this one using your hands-on experience.

The apply_along_axis function is a NumPy function that allows you to apply a function to each row or column of a NumPy array. This can be useful if you want to perform some operation on each row or column of the array and don't want to use a loop.

The syntax for using apply_along_axis is as follows:

np.apply_along_axis(func, axis, arr, *args, **kwargs) 

• func: This is the function that will be applied to each row or column of the array. It should be a function that takes a 1D array as input and returns a scalar value.
• axis: This specifies the axis along which the function will be applied. If axis=0, the function will be applied to each column of the array. If axis=1, the function will be applied to each row of the array.
• arr: This is the NumPy array on which the function will be applied.
• *args: Any additional arguments to the function can be passed using the *args syntax. These arguments will be passed to the function unchanged.
• **kwargs: Any additional keyword arguments to the function can be passed using the **kwargs syntax. These arguments will be passed to the function as keyword arguments.

Here is an example of how to use apply_along_axis to apply a function to each row of a NumPy array:

import numpy as np 
# Define a function that takes a 1D array and returns the sum of its elements 
def sum_elements(x): 
return np.sum(x) 
# Create a NumPy array 
data = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) 
# Apply the function to each row of the array 
result = np.apply_along_axis(sum_elements, axis=1, arr=data) 
print(result) # prints [6 15 24] 

This will apply the sum_elements function to each row of the array data, and return a new array containing the results. The resulting array will be printed to the console.

You can also use apply_along_axis to apply a function to each column of a NumPy array by setting axis=0. For example:

import numpy as np 

# Define a function that takes a 1D array and returns the sum of its elements

def sum_elements(x): 
return np.sum(x) 
# Create a NumPy array 
data = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]]) 

# Apply the function to each column of the array

result = np.apply_along_axis(sum_elements, axis=0, arr=data) 
print(result) # prints [12 15 18] 

This will apply the sum_elements function to each column of the array data and return a new array containing the results. The resulting array will be printed to the console.

The fast Fourier transform (FFT) is an efficient algorithm for computing the discrete Fourier transform (DFT) of a sequence. The DFT is a mathematical operation that decomposes a sequence of values into its component frequencies. This can be useful for analyzing the frequency content of a signal, such as a time series or an audio signal. 

NumPy's fft module provides functions for performing FFTs on NumPy arrays. The fft.fft function is the main function for computing FFTs. It takes a NumPy array as input and returns the FFT of the array as a NumPy array of complex numbers. The fft.fftfreq function is used to generate the frequencies corresponding to the FFT coefficients. 

Here's an example of how to use these functions to compute and plot the FFT of a 1D NumPy array: 

import numpy as np 
import matplotlib.pyplot as plt 
# Generate a test signal with four sine waves at different frequencies 
t = np.linspace(0, 2*np.pi, 1000, endpoint=False) 
sig = np.sin(2*t) + np.sin(6*t) + np.sin(10*t) + np.sin(14*t) 
# Compute the FFT of the signal 
sig_fft = np.fft.fft(sig) 
# Get the frequencies corresponding to the FFT coefficients 
frequencies = np.fft.fftfreq(sig.size, t[1] - t[0]) 
# Only keep the positive frequencies 
positive_freqs = frequencies[:sig.size // 2] 
sig_fft = sig_fft[:sig.size // 2] 
# Plot the FFT 
plt.plot(positive_freqs, np.abs(sig_fft)) 
plt.xlabel('Frequency (Hz)') 
plt.ylabel('FFT Coefficient') 
plt.show() 

Jupyter Notebook: https://github.com/rajshashwatcodes/KnowledgeHut/blob/main/NumpyInterviewQuestions/NumpyAdvance11a.ipynb 

This code generates a test signal that consists of four sine waves at different frequencies, then computes the FFT of the signal using the fft.fft function. The fft.fftfreq function is used to generate the frequencies corresponding to the FFT coefficients, and the positive frequencies are extracted from the resulting array. Finally, the FFT coefficients are plotted as a function of frequency. 

The fft.fft function can also be used to perform FFTs on 2D NumPy arrays, by specifying the axis parameter. For example, to compute the FFT of each column in a 2D array, you can set axis=0. 

The fft function can also be used to perform FFTs on 2D NumPy arrays, by applying the FFT to each row or column of the array. For example: 

import numpy as np 
# Generate a test signal with four sine waves at different frequencies 
t = np.linspace(0, 2*np.pi, 1000, endpoint=False) 
sig = np.sin(2*t) + np.sin(6*t) + np.sin(10*t) + np.sin(14*t) 
# Add some noise to the signal 
sig += 0.1 * np.random.randn(sig.size) 
# Reshape the signal into a 2D array with 10 rows and 100 columns 
sig_2d = sig.reshape((10, 100)) 
# Compute the FFT of each column 
sig_fft = np.fft.fft(sig_2d, axis=0) 
# Get the frequencies corresponding to the FFT coefficients 
frequencies = np.fft.fftfreq(sig_2d.shape[1], t[1] - t[0]) 
# Only keep the positive frequencies 
positive_freqs = frequencies[:sig_2d.shape[1] // 2] 
sig_fft = sig_fft[:sig_2d.shape[1] // 2, :] 
# Plot the FFT for each row 
plt.imshow(np.abs(sig_fft), extent=(positive_freqs[0], positive_freqs[-1], sig_2d.shape[0], 0)) 
plt.xlabel('Frequency (Hz)') 
plt.ylabel('Row') 
plt.colorbar() 

To find the local peaks (or maxima) in a 1-D NumPy array, you can use the NumPy.where function along with the NumPy.greater function to create a Boolean mask indicating the positions of the local peaks. 

First, we compute the differences between adjacent elements in the array using np.diff. This is done by taking the slice of the array arr[1:] and subtracting it from the slice arr[:-1]. 

import numpy as np 
arr = np.array([1, 2, 3, 2, 1, 2, 3, 4, 3, 2, 1]) 
diff = np.diff(arr) 
print(diff) 

This will output the differences between adjacent elements: [1 1 -1 -1 1 1 1 -1 -1 1] 

Next, we use the NumPy.greater function to create a Boolean mask indicating the positions where the differences are greater than 0. This will give us a mask for the rising edges of the peaks in the array. 

rising_edges = np.greater(diff[:-1], 0) 
print(rising_edges) 

This will output a Boolean mask for the rising edges: [ True True False False True True True False False True] 

We can use the same method to create a Boolean mask for the falling edges of the peaks by comparing the differences to 0. 

falling_edges = np.greater(diff[1:], 0) 
print(falling_edges) 

This will output a Boolean mask for the falling edges: [False True True True False True True True True False] 

Finally, we use the NumPy.where function to find the indices where both masks are True, indicating the positions of the local maxima. We use the & operator to compute the element-wise logical AND of the two masks. 

maxima_mask = rising_edges & falling_edges 
maxima_indices = np.where(maxima_mask)[0] + 1 
print(maxima_indices) 

This will output the indices of the local maxima: [2, 6, 8] 

Or we can use NumPy.argmax and NumPy.maximum 

First, we use the NumPy.argmax function to find the indices of the maximum element in the input array. We set the axis parameter to None, which will flatten the input array and find the maximum element. 

import numpy as np 
arr = np.array([1, 2, 3, 2, 1, 2, 3, 4, 3, 2, 1]) 
maxima_indices = np.argmax(arr) 
print(maxima_indices) 

This will output the index of the maximum element: 7 

Next, we use the NumPy.maximum function to create a mask of the local maxima by comparing the elements of the input array to the maximum element found by NumPy.argmax. 

maxima_mask = np.maximum(arr) 
print(maxima_mask) 

This will output a mask of the local maxima: [False False False False False False True True True False False] 

Finally, we use the NumPy.where function to find the indices where the mask is True, indicating the positions of the local maxima. 

maxima_indices = np.where(maxima_mask)[0] 
print(maxima_indices) 

This will output the indices of the local maxima: [6, 7, 8] 

SWIG is a tool that is used to generate language bindings for C and C++ code. It works by taking the C or C++ header files that define the functions and methods you want to expose to other languages, and generating wrapper code that can be used to call these functions and methods from other languages. 

NumPy provides a set of functions and methods for performing mathematical operations on arrays and matrices, and SWIG can be used to expose these functions and methods to Python so that they can be used in Python programs. 

For example, suppose you have a C library with a function called add that takes two integers as arguments and returns their sum. You can use SWIG to generate Python bindings for this function, which will allow you to call the add function from a Python program. 

To do this, you would create a SWIG interface file that describes the functions and methods you want to expose to Python. This file will typically have a .i extension and will contain directives that tell SWIG how to generate the wrapper code. 

For example, the SWIG interface file for the add function might look like this: 

%module example 
int add(int x, int y); 

You can then run SWIG on this interface file to generate the wrapper code. The wrapper code will typically be a C or C++ file with a _wrap.c or _wrap.cpp extension. 

To use the wrapper code in a Python program, you can import it using NumPy's ctypeslib module. This module provides a set of functions for loading and using C libraries in Python programs. 

For example, you can use the ctypeslib.load_library function to load the C library and the generated wrapper code, and then call the add function from Python like this: 

import numpy as np 
import ctypeslib 
# Load the C library and the generated wrapper code using ctypes 
lib = ctypes.cdll.LoadLibrary('path/to/library.so') 
bindings = ctypeslib.load_library('path/to/library', 'path/to/library_wrap.c') 
# Call the C function from Python 
result = bindings.add(1, 2) 
print(result) # prints 3 

Here is another example of using SWIG and NumPy's ctypeslib module to call a C function from a Python program: 

import numpy as np 
import ctypeslib 
# Load the C library and the generated wrapper code using ctypes 
lib = ctypes.cdll.LoadLibrary('path/to/library.so') 
bindings = ctypeslib.load_library('path/to/library', 'path/to/library_wrap.c') 
# Define the C function signature using ctypes 
bindings.add.argtypes = [ctypes.c_int, ctypes.c_int] 
bindings.add.restype = ctypes.c_int 
# Call the C function from Python 
result = bindings.add(1, 2) 
print(result) # prints 3 

In this example, we use the argtypes and restype attributes of the add function to specify the argument and return types of the C function. This is necessary because NumPy's ctypeslib module does not provide type information for the C functions, and we need to specify the types explicitly using ctypes. 

NumPy provides several options for handling numerical exceptions, such as : 

• Overflow: An overflow error occurs when the result of a computation is too large to be represented by the data type being used. For example, if you try to compute the square root of a negative number using the NumPy.sqrt function, NumPy will raise a ValueError exception. 
• Underflow: An underflow error occurs when the result of a computation is too small to be represented by the data type being used. For example, if you try to compute the reciprocal of a very large number using the NumPy.reciprocal function, NumPy will raise a FloatingPointError exception. 
• Divide-by-zero: A divide-by-zero error occurs when you try to divide a number by zero. For example, if you try to divide 5 by 0 using the NumPy.divide function, NumPy will raise a FloatingPointError exception. 
• Invalid: An invalid error occurs when a computation produces an invalid result, such as a complex number or nan (not a number). For example, if you try to compute the square root of a negative number using the NumPy.sqrt function, NumPy will raise a FloatingPointError exception. 

You can use the NumPy.seterr function to specify the error behavior for four types of exceptions: overflow, underflow, divide-by-zero, and invalid. The seterr function takes four parameters, one for each type of exception, and each parameter can have one of three values:  

• 'ignore': NumPy will ignore the exception and continue executing the code. 
• 'warn': NumPy will print a warning message to the console and continue executing the code. 
• 'raise': NumPy will raise an exception and halt execution. 

Here is an example of using the NumPy.seterr function to specify the error behavior for different types of exceptions: 

import numpy as np 

# Ignore overflow and underflow errors, print a warning for divide-by-zero errors, and raise an exception for invalid errors 

np.seterr(overflow='ignore', underflow='ignore', divide='warn', invalid='raise') 

By default, NumPy will raise an exception when it encounters a numerical error, such as an overflow or underflow. For example, if you try to compute the square root of a negative number using the NumPy.sqrt function, NumPy will raise a ValueError exception. 

You can also use the NumPy.seterr function to specify how NumPy should handle numerical exceptions. The seterr function allows you to set the error behavior for any of four types of exceptions: overflow, underflow, divide-by-zero, and invalid. 

For example, you can use the following code to ignore overflow and underflow errors: 

import numpy as np 
np.seterr(overflow='ignore', underflow='ignore') 

You can also use the NumPy.seterr function to specify that NumPy should raise an exception when it encounters a numerical error. For example, you can use the following code to raise an exception for divide-by-zero errors: 

import numpy as np 
np.seterr(divide='raise') 

If you do not specify the error behavior using the NumPy.seterr function, NumPy will use the default behavior, which is to raise an exception for all types of errors. 

The meshgrid function in NumPy is a tool for creating a grid of coordinates from two or more one-dimensional coordinate arrays. It takes as input a set of 1D arrays representing the coordinates along each dimension and returns a set of ND arrays representing the coordinates at each point in the grid. 

For example, consider the following code: 

import numpy as np 
# Create 1D coordinate arrays 
x = np.array([1, 2, 3]) 
y = np.array([4, 5, 6]) 
# Create a grid of coordinates using meshgrid 
X, Y = np.meshgrid(x, y) 
print(X) # prints [[1 2 3] 
# [1 2 3] 
# [1 2 3]] 
print(Y) # prints [[4 4 4] 
# [5 5 5] 
# [6 6 6]] 

The meshgrid function returns two 2D arrays, X and Y, which represent the coordinates of a 3x3 grid. The first array, X, contains the x-coordinates of the grid points, and the second array, Y, contains the y-coordinates. 

You can use the meshgrid function to evaluate functions on a grid, plot data on a grid, etc. It is a useful tool for working with multidimensional data in NumPy. 

The meshgrid function is used to create a grid of coordinates from two or more one-dimensional coordinate arrays. It is useful for a variety of tasks, such as evaluating functions on a grid, plotting data on a grid, and working with multidimensional data. 

For example, you can use the meshgrid function to plot a 3D surface: 

import numpy as np 
import matplotlib.pyplot as plt 
# Create 1D coordinate arrays 
x = np.linspace(-2, 2, 100) 
y = np.linspace(-2, 2, 100) 
# Create a grid of coordinates using meshgrid 
X, Y = np.meshgrid(x, y) 
# Evaluate a 3D function on the grid 
Z = X**2 - Y**2 
# Plot the 3D surface 
fig = plt.figure() 
ax = fig.add_subplot(111, projection='3d') 
ax.plot_surface(X, Y, Z) 
plt.show() 

This code creates a 100x100 grid of coordinates using the meshgrid function, and then evaluates a 3D function (X**2 - Y**2) on the grid. The resulting 3D surface is plotted using Matplotlib. 

The meshgrid function is also useful for working with multidimensional data, such as images. For example, you can use it to create a grid of coordinates that can be used to index into an image array: 

import numpy as np 
# Create 1D coordinate arrays 
x = np.arange(10) 
y = np.arange(10) 
# Create a grid of coordinates using meshgrid 
X, Y = np.meshgrid(x, y) 
# Create a random image 
image = np.random.random((10, 10)) 
# Use the grid of coordinates to index into the image 
pixel_values = image[X, Y] 

This code creates a 10x10 grid of coordinates using the meshgrid function, and then uses the grid to index into a random image array. The resulting array pixel_values contains the pixel values at each point in the grid. 

The “ndim” attribute in NumPy is an attribute of the ndarray class that returns the number of dimensions (axes) of the array. It is a property of the array, not a function, so you do not need to call it with parentheses. 

For example, consider the following code: 

import numpy as np 
# Create a 1D array 
a = np.array([1, 2, 3]) 
print(a.ndim) # prints 1 
# Create a 2D array 
b = np.array([[1, 2, 3], [4, 5, 6]]) 
print(b.ndim) # prints 2 
# Create a 3D array 
c = np.array([[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [10, 11, 12]]]) 
print(c.ndim) # prints 3 

In this example, the ndim attribute returns the number of dimensions of the array. The 1D array a has ndim equal to 1, the 2D array b has ndim equal to 2, and the 3D array c has ndim equal to 3.For example, consider the following code: 

import numpy as np 
# Create a 1D array 
a = np.array([1, 2, 3]) 
print(a.ndim) # prints 1 
# Create a 2D array 
b = np.array([[1, 2, 3], [4, 5, 6]]) 
print(b.ndim) # prints 2 
# Create a 3D array 
c = np.array([[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [10, 11, 12]]]) 
print(c.ndim) # prints 3 

In this example, the ndim attribute returns the number of dimensions of the array. The 1D array a has ndim equal to 1, the 2D array b has ndim equal to 2, and the 3D array c has ndim equal to 3. 

The ndim attribute is useful for determining the shape of an array, which can be useful for indexing into the array correctly. For example: 

import numpy as np 
# Create a 3D array 
a = np.array([[[1, 2, 3], [4, 5, 6]], [[7, 8, 9], [10, 11, 12]]]) 
# Get the shape of the array 
shape = a.shape 
print(shape) # prints (2, 2, 3) 
# Index into the array using the shape 
x, y, z = a[0, 0, 0], a[1, 1, 1], a[shape[0]-1, shape[1]-1, shape[2]-1] 
print(x, y, z) # prints 1 11 12 

In this example, the shape of the 3D array a is (2, 2, 3), indicating that it has 2 elements along the first dimension, 2 elements along the second dimension, and 3 elements along the third dimension. We can use the shape of the array to index into it correctly, as shown in the example. 

You can also use the ndim attribute to iterate over the elements of an array. For example: 

import numpy as np 
# Create a 2D array 
a = np.array([[1, 2, 3], [4, 5, 6]]) 
# Iterate over the elements of the array 
for i in range(a.ndim): 
for j in range(a.shape[i]): 
print(a[i, j]) 

# Output: 

# 1 
# 2 
# 3 
# 4 
# 5 
# 6 

There is no one "best" way to create a histogram, as the appropriate method will depend on your specific needs and the context in which the histogram will be used. Here are a few common ways to create histograms: 

• Using Matplotlib: Matplotlib is a popular Python library for creating a wide variety of plots and charts. To create a histogram with Matplotlib, you can use the hist function, which takes in a sequence of data and returns histogram bin counts and edges. You can then use the plot function to visualize the histogram. 
• Using Seaborn: Seaborn is a library built on top of Matplotlib that provides a high-level interface for creating many different types of plots, including histograms. To create a histogram with Seaborn, you can use the histplot function, which takes in a Pandas DataFrame and plots a histogram of one of the columns. 
• Using NumPy: NumPy is a library for working with numerical data in Python. It includes a function called histogram that can be used to compute histograms of data. The histogram function returns the histogram bin counts and edges, but does not provide a way to visualize the histogram. You can use Matplotlib or Seaborn to visualize the histogram using the returned bin counts and edges. 
• Using Pandas: Pandas is a library for working with data in Python. It includes a function called plot.hist that can be used to create histograms of data in a Pandas DataFrame. The plot.hist function uses Matplotlib to visualize the histogram. 

Overall, the choice of which method to use will depend on your specific needs and the tools that you are comfortable using. Matplotlib and Seaborn are both popular choices for creating histograms, but NumPy and Pandas can also be useful depending on the context. To compute histograms of data using NumPy's histogram function, you will need to pass it the following arguments: 

• a: This should be a sequence of data, such as a list or NumPy array. The histogram will be computed for the values in this sequence. 
• bins: This specifies the number of bins that you want to use to divide up the range of data. You can either specify the number of bins directly, or you can pass a sequence of bin edges (e.g., [0, 10, 20, 30]). 
• range: This specifies the range of data that you want to include in the histogram. It should be a tuple of the form (min, max), where min is the minimum value to include and max is the maximum value to include. 

For example, to compute a histogram of data in the range [0, 10] with 10 bins, you could do the following: 

import numpy as np 
# Generate some random data 
data = np.random.uniform(0, 10, 1000) 
# Compute the histogram 
hist, bins = np.histogram(data, bins=10, range=(0, 10)) 

The hist variable will contain the histogram counts, and the bins variable will contain the bin edges. You can then use these values to visualize the histogram using Matplotlib or some other library. 

In NumPy, a stride is a tuple of indices that specifies how to index into an array. The stride for an array specifies the number of indices that you need to skip in order to move to the next element in a particular dimension. 

For example, consider the following 2D array: 

import numpy as np 
a = np.array([[1, 2, 3], [4, 5, 6]]) 

This array has shape (2, 3), which means it has 2 rows and 3 columns. The stride for this array specifies how many indices you need to skip in order to move to the next element in each dimension. The stride for the first dimension (rows) will be the number of elements in a single row, while the stride for the second dimension (columns) will be the number of bytes in a single element. 

For example, the stride for the first dimension (rows) will be 3, since there are 3 elements in a row. The stride for the second dimension (columns) will depend on the data type of the array. If the array has a data type of int32, for example, the stride for the second dimension will be 4, since int32 values are 4 bytes each. 

You can access the strides of an array using the strides attribute. For example: 

import numpy as np 
a = np.array([[1, 2, 3], [4, 5, 6]]) 
print(a.strides) 

This will output (12, 4), which indicates that to move to the next element in the first dimension, you need to skip 12 indices, and to move to the next element in the second dimension, you need to skip 4 indices. 

You can also use the as_strided function to create a new array with a specified stride. This can be useful if you want to create a view of an array with a different stride than the original array. 

Yes, it is possible to create strides from a 1D array in NumPy. A stride is a tuple of indices that specifies how to index into an array. The stride for an array specifies the number of indices that you need to skip in order to move to the next element in a particular dimension. 

To create strides from a 1D array in NumPy, you can use the strides attribute of the array. This attribute returns a tuple of strides, one for each dimension of the array. For a 1D array, the strides tuple will contain a single element. 

Here is an example of how to create strides from a 1D array in NumPy: 

import numpy as np 
# Create a 1D array 
a = np.array([1, 2, 3, 4, 5]) 
# Print the strides of the array 
print(a.strides) 

This will output (4,), which indicates that to move to the next element in the array, you need to skip 4 indices (since the data type of the array is int32, which has a size of 4 bytes). 

You can also use the as_strided function to create a new array with a specified stride. This can be useful if you want to create a view of an array with a different stride than the original array. For example: 

import numpy as np 
# Create a 1D array 
a = np.array([1, 2, 3, 4, 5]) 
# Create a view of the array with a stride of 2 
b = np.lib.stride_tricks.as_strided(a, shape=(3,), strides=(8,)) 
# Print the strides of the new array 
print(b.strides) 

This will output (8,), indicating that the new array has a stride of 8

The asanyarray is a function in the NumPy library that converts an object to a NumPy array, while preserving the subclass type of the object if it is already a NumPy array. 

For example, consider the following code: 

import numpy as np 
# Define a custom subclass of ndarray 
class MyArray(np.ndarray): 
def __new__(cls, data): 
# Create a new ndarray instance 
obj = np.asarray(data).view(cls) 
return obj 
# Create an instance of MyArray 
a = MyArray([1, 2, 3]) 
# Convert the instance to a NumPy array using asanyarray 
b = np.asanyarray(a) 

In this example, a is an instance of the MyArray subclass of ndarray. When a is passed to asanyarray, it returns a new NumPy array that is a copy of a, but with the same subclass type (i.e., MyArray). 

The asanyarray is similar to the array function, but it allows you to preserve the subclass type of an array if it is already a NumPy array. array always returns a new NumPy array, regardless of the input type. 

The asanyarray can be useful when you want to ensure that an object is a NumPy array, but you want to preserve any additional functionality or attributes that may be defined in a subclass of ndarray. 

To use NumPy's asanyarray function to convert objects to NumPy arrays while preserving their subclass type, you can pass the object to asanyarray as an argument. asanyarray will then attempt to convert the object to a NumPy array, and if the object is already a NumPy array, it will return the array without making a copy. 

Here is an example of how to use asanyarray to convert an object to a NumPy array while preserving its subclass type: 

import numpy as np 
# Define a custom subclass of ndarray 
class MyArray(np.ndarray): 
def __new__(cls, data): 
# Create a new ndarray instance 
obj = np.asarray(data).view(cls) 
return obj 
# Create an instance of MyArray 
a = MyArray([1, 2, 3]) 
# Convert the instance to a NumPy array using asanyarray 
b = np.asanyarray(a) 
# Print the type of the resulting array 
print(type(b)) 

This will output , indicating that the resulting array is an instance of the MyArray subclass. 

Keep in mind that asanyarray only works on objects that can be converted to NumPy arrays using asarray. If the object cannot be converted using asarray, asanyarray will raise a TypeError. 

A staple in NumPy advanced interview questions and answers, be prepared to answer this one using your hands-on experience.

NumPy provides support for parallel computation using the NumPy.distributed module. This module provides functions for distributing large arrays across multiple CPU cores or even multiple machines.

To use NumPy's support for parallel computation, you will first need to install the dask library. Dask is a parallel computing library that NumPy uses to distribute work across multiple CPU cores or machines.

Once you have dask installed, you can use the NumPy.distributed module to perform operations on large arrays using multiple CPU cores. Here is an example of how to use NumPy's distributed arrays to calculate the sum of a large array using multiple CPU cores:

import numpy as np 
import dask.array as da 
# Create a large array using dask 
x = da.random.random(size=(10000, 10000), chunks=(1000, 1000)) 
# Calculate the sum of the array using multiple CPU cores 
result = np.sum(x) 
# Print the result 
print(result) 

In this example, x is a large array created using dask.array. The chunks parameter specifies the size of the chunks that the array should be divided into for parallel processing. When the sum function is called on x, NumPy will use multiple CPU cores to calculate the sum in parallel.

You can also use the NumPy.distributed.Client class to specify the number of CPU cores to use for parallel computation. For example:

import numpy as np 
import dask.array as da 
from dask.distributed import Client 
# Start a dask client with 4 CPU cores 
client = Client(n_workers=4) 
# Create a large array using dask 
x = da.random.random(size=(10000, 10000), chunks=(1000, 1000)) 
# Calculate the sum of the array using 4 CPU cores 
result = np.sum(x) 
# Print the result 
print(result) 
# Shut down the dask client 
client.close() 

In this example, the Client class is used to start a dask client with 4 CPU cores. The NumPy.sum function is then called on the distributed array x, and the sum is calculated using 4 CPU cores in parallel.