📊 Day 21 : NumPy Fundamentals

🎯 Enterprise Objective

Welcome to Phase 2. Data Analytics requires crunching millions of numbers. Standard Python is too slow. NumPy bridges Python with fast C-code, providing the N-Dimensional Array (ndarray). Mastering vectorized math is the gateway to Pandas and Machine Learning.

📋 Strategic Overview

#	Topic	Concept
1	ndarrays	Homogeneous matrices
2	Generation	`arange`, `zeros`, `linspace`
3	Vectorization	Fast math without loops

1. NumPy Arrays (ndarrays) : The Core of Data Science

🔍 What is it?

Welcome to Phase 2: Data Analytics. NumPy (Numerical Python) is the foundation of Pandas and Scikit-Learn. Its core structure is the ndarray (N-dimensional array), which is up to 50x faster than Python lists because it uses fixed-type C-arrays under the hood.

import numpy as np
# Create a 1D array
arr = np.array([1, 2, 3, 4, 5])

💼 Why Data Analysts Care

• Performance: Processing millions of data points instantly

• Memory Efficiency: NumPy arrays take up significantly less RAM than standard Python lists

⚠️ Mixed Data Types

Unlike Python lists which can hold [1, 'a', True], a NumPy array must be homogenous (all elements must be the same type, e.g., all floats or all ints). If you mix them, NumPy will force them into strings.

In [ ]:

🧪 Concept Checks: `NumPy Basics`

Q1. Import numpy as np. Create an array from [10, 20, 30]. Print it.

In [ ]:

Q2. Create a 2D array (matrix) representing a 3x3 grid of numbers. Print its .shape.

In [ ]:

Q3. Try creating an array from [1, "two", 3]. Print its .dtype. Notice how the numbers became strings ().



In [ ]:
# Write your answer here

Q4. Print the number of dimensions .ndim of the 3x3 array.

In [ ]:
# Write your answer here

Q5. Print the total number of elements in the array using .size.

In [ ]:
# Write your answer here




    2. Array Generation : Built-in Constructors


    🔍 What is it?


You rarely build massive arrays by manually typing lists. NumPy provides powerful generation functions like np.zeros(), np.arange(), and np.linspace().

Function Purpose Example

np.zeros(shape) Array of 0s np.zeros((3,3))
np.ones(shape) Array of 1s np.ones(5)
np.arange(start, stop, step) Like Python range np.arange(0, 10, 2)
np.linspace(start, stop, num) Evenly spaced points np.linspace(0, 1, 5)

💼 Why Data Analysts Care
• Initialization: Creating empty matrices to fill with model weights
• Plotting: Using linspace to generate perfectly spaced X-axis values for charts
🧠 Pro Tip

    np.arange() works with floats (e.g., step=0.5), unlike Python's built-in range() which only takes integers!


In [ ]:
import numpy as np

# Zeros and Ones
zeros = np.zeros((2, 4))  # Pass shape as a tuple
print("Zeros:\n", zeros)

# Arange (Start, Stop exclusive, Step)
steps = np.arange(0, 10, 2)
print("\nArange:", steps)

# Linspace (Start, Stop INCLUSIVE, Number of points)
points = np.linspace(0, 100, 5)
print("\nLinspace:", points)

# Random generation (Uniform 0-1)
rands = np.random.rand(3, 3)
print("\nRandoms:\n", rands)




    🧪 Concept Checks: Generators

Q1. Create an array of 10 zeros. Print it.

In [ ]:
# Write your answer here

Q2. Create a 3x3 matrix of ones using np.ones((3, 3)).

In [ ]:
# Write your answer here

Q3. Use np.arange() to create an array of numbers from 10 to 50, stepping by 5.

In [ ]:
# Write your answer here

Q4. Use np.linspace() to generate exactly 11 evenly spaced points between 0 and 1.

In [ ]:
# Write your answer here

Q5. Generate a 2x2 matrix of random integers between 1 and 100 using np.random.randint(1, 100, size=(2,2)).

In [ ]:
# Write your answer here




    3. Vectorized Operations : No More For-Loops


    🔍 What is it?


The most important concept in NumPy is Vectorization. You can perform math on entire arrays at once without writing for loops. This is executed in optimized C code, making it blazing fast.
# ❌ Bad (Python Loop)
for i in range(len(arr)): arr[i] *= 2

# ✅ Good (Vectorized)
arr = arr * 2
💼 Why Data Analysts Care
• Financial Math: Applying a 5% interest rate to a million bank accounts instantly (balances * 1.05)
• Image Processing: Brightening an image matrix by adding a constant (pixels + 50)
🧠 Pro Tip


Arrays operate element-wise. If you add two arrays A + B, they must be the exact same shape (or compatible via 'broadcasting'). They add position 0 to position 0, 1 to 1, etc.

In [ ]:
import numpy as np
import time

arr = np.array([1, 2, 3, 4, 5])

# Scalar Math (Applies to all elements)
print(f"Add: {arr + 10}")
print(f"Multiply: {arr * 10}")
print(f"Power: {arr ** 2}")

# Array to Array Math (Element-wise)
arr2 = np.array([10, 20, 30, 40, 50])
print(f"\nArray Addition: {arr + arr2}")

# Speed test!
big_list = list(range(1000000))
big_arr = np.array(big_list)

t1 = time.time()
_ = [x * 2 for x in big_list]  # Python loop
list_time = time.time() - t1

t2 = time.time()
_ = big_arr * 2  # NumPy Vectorization
arr_time = time.time() - t2

print(f"\nNumPy is {list_time / arr_time:.1f}x faster!")




    🧪 Concept Checks: Vectorization

Q1. Create arr = np.array([10, 20, 30]). Divide every element by 10 and print the result.

In [ ]:
# Write your answer here

Q2. Given A = np.array([1, 2]) and B = np.array([10, 20]), multiply them together. Print the result.

In [ ]:
# Write your answer here

Q3. Create an array using arange(1, 6). Square all elements (**2). Print the result.

In [ ]:
# Write your answer here

Q4. Write a boolean operation: arr > 15. Print the result (you get a boolean array!).

In [ ]:
# Write your answer here

Q5. Measure the speed difference between sum(range(100000)) and np.arange(100000).sum().

In [ ]:
# Write your answer here



🛠️ Professional Practice Tasks
Theory is useless without muscle memory. Complete these tasks to solidify your understanding.
Task 1 (Celsius to Fahrenheit): Create an array of Celsius temperatures: [0, 10, 20, 30, 40]. Use vectorized math to convert them to Fahrenheit (C * 9/5) + 32. Print the result.

In [ ]:
# Write your answer here

Task 2 (Distance Formula): Given two 1D arrays p1 = np.array([1, 2, 3]) and p2 = np.array([4, 5, 6]). Calculate the Euclidean distance using vector math: sqrt(sum((p1 - p2)**2)). (Use np.sqrt and .sum()).

In [ ]:
# Write your answer here

Task 3 (Identity Matrix): Research and use np.eye() to create a 5x5 Identity matrix (1s on the diagonal, 0s elsewhere). Multiply it by 5 to make the diagonal 5s.

In [ ]:
# Write your answer here

Task 4 (Random Noise): Create a 100-element array using np.linspace(0, 10, 100). Add random uniform noise to it using np.random.rand(100). Plotting this is the foundation of data visualization!

In [ ]:
# Write your answer here

Task 5 (Shape Manipulation): Create a 1D array of 12 elements. Use .reshape((3, 4)) to turn it into a 3x4 matrix. Print the new matrix and its .shape.

In [ ]:
# Write your answer here



💻 Pure Coding Interview Questions


Q1.
Why is NumPy so much faster than standard Python lists?


In [ ]:
# Solution for Q1



Q2.
What is the difference between an ndarray and a Python list?


In [ ]:
# Solution for Q2



Q3.
Explain what 'Vectorization' means in NumPy.


In [ ]:
# Solution for Q3



Q4.
What happens if you try to put a string into an array of integers? Explain dtype casting.


In [ ]:
# Solution for Q4



Q5.
What does the .shape attribute return? What type of object is it?


In [ ]:
# Solution for Q5



Q6.
How do you create a 3D array in NumPy? What does its shape look like?


In [ ]:
# Solution for Q6



Q7.
Explain the difference between np.arange() and np.linspace().


In [ ]:
# Solution for Q7



Q8.
What is 'Element-wise' operation?


In [ ]:
# Solution for Q8



Q9.
How does NumPy handle missing data (NaN) compared to Python None?


In [ ]:
# Solution for Q9



Q10.
Write a one-liner to generate an array of 100 random numbers drawn from a normal distribution (np.random.randn).


In [ ]:
# Solution for Q10



Q11.
What is the output of np.array([1, 2]) + np.array([3, 4, 5])? Explain what happens.


In [ ]:
# Solution for Q11



Q12.
How do you check the data type of a NumPy array?


In [ ]:
# Solution for Q12



Q13.
How do you explicitly force a NumPy array to be floats instead of ints upon creation? (Hint: dtype=float).


In [ ]:
# Solution for Q13



Q14.
Explain what np.zeros_like(arr) does.


In [ ]:
# Solution for Q14



Q15.
What is the difference between np.random.rand() and np.random.randint()?


In [ ]:
# Solution for Q15



Q16.
How do you find the total memory consumed by a NumPy array? (Hint: .nbytes).


In [ ]:
# Solution for Q16



Q17.
Why is it a bad idea to use append() in a loop with NumPy arrays?


In [ ]:
# Solution for Q17



Q18.
What is broadcasting in NumPy (briefly)?


In [ ]:
# Solution for Q18



Q19.
How do you flatten a 2D matrix into a 1D array? (.flatten() or .ravel()).


In [ ]:
# Solution for Q19



Q20.
Explain the difference between deep copy and shallow copy (views) in NumPy arrays.


In [ ]:
# Solution for Q20



Q21.
What happens when you do arr > 5? What does it return?


In [ ]:
# Solution for Q21



Q22.
How do you get the transpose of a matrix in NumPy? (Hint: .T).


In [ ]:
# Solution for Q22



Q23.
What is the dot product of two vectors, and how do you compute it in NumPy? (np.dot or @).


In [ ]:
# Solution for Q23



Q24.
How do you find the maximum value in an array? How do you find its index? (argmax).


In [ ]:
# Solution for Q24



Q25.
What is the difference between np.nan and np.inf?


In [ ]:
# Solution for Q25



📊 Day 21 Executive Summary

# Topic Key Takeaway

1 Array shape, ndim, and dtype define the structure
2 Gen Never build big arrays by hand
3 Vector Replace for loops with arr * 2. It's 100x faster.

✅ Instructor's End-of-Day Checklist
• [ ] I can create 1D and 2D arrays.
• [ ] I can generate arrays using linspace and arange.
• [ ] I understand vectorized math vs for-loops.

                🚀 Continue to Day 22 - NumPy Advanced: Indexing, Filtering & Broadcasting

Function	Purpose	Example
`np.zeros(shape)`	Array of 0s	`np.zeros((3,3))`
`np.ones(shape)`	Array of 1s	`np.ones(5)`
`np.arange(start, stop, step)`	Like Python `range`	`np.arange(0, 10, 2)`
`np.linspace(start, stop, num)`	Evenly spaced points	`np.linspace(0, 1, 5)`

#	Topic	Key Takeaway
1	Array	`shape`, `ndim`, and `dtype` define the structure
2	Gen	Never build big arrays by hand
3	Vector	Replace `for` loops with `arr * 2`. It's 100x faster.

📊 Day 21 : NumPy Fundamentals

🎯 Enterprise Objective

📋 Strategic Overview

1. NumPy Arrays (ndarrays) : The Core of Data Science

💼 Why Data Analysts Care

⚠️ Mixed Data Types

🧪 Concept Checks: NumPy Basics

2. Array Generation : Built-in Constructors

💼 Why Data Analysts Care

🧠 Pro Tip

🧪 Concept Checks: Generators

3. Vectorized Operations : No More For-Loops

💼 Why Data Analysts Care

🧠 Pro Tip

🧪 Concept Checks: Vectorization

🛠️ Professional Practice Tasks

💻 Pure Coding Interview Questions

Why is NumPy so much faster than standard Python lists?

What is the difference between an ndarray and a Python list?

Explain what 'Vectorization' means in NumPy.

What happens if you try to put a string into an array of integers? Explain dtype casting.

What does the .shape attribute return? What type of object is it?

How do you create a 3D array in NumPy? What does its shape look like?

Explain the difference between np.arange() and np.linspace().

What is 'Element-wise' operation?

How does NumPy handle missing data (NaN) compared to Python None?

Write a one-liner to generate an array of 100 random numbers drawn from a normal distribution (np.random.randn).

What is the output of np.array([1, 2]) + np.array([3, 4, 5])? Explain what happens.

How do you check the data type of a NumPy array?

How do you explicitly force a NumPy array to be floats instead of ints upon creation? (Hint: dtype=float).

Explain what np.zeros_like(arr) does.

What is the difference between np.random.rand() and np.random.randint()?

How do you find the total memory consumed by a NumPy array? (Hint: .nbytes).

Why is it a bad idea to use append() in a loop with NumPy arrays?

What is broadcasting in NumPy (briefly)?

How do you flatten a 2D matrix into a 1D array? (.flatten() or .ravel()).

Explain the difference between deep copy and shallow copy (views) in NumPy arrays.

What happens when you do arr > 5? What does it return?

How do you get the transpose of a matrix in NumPy? (Hint: .T).

What is the dot product of two vectors, and how do you compute it in NumPy? (np.dot or @).

How do you find the maximum value in an array? How do you find its index? (argmax).

What is the difference between np.nan and np.inf?

📊 Day 21 Executive Summary

✅ Instructor's End-of-Day Checklist

🧪 Concept Checks: `NumPy Basics`

🧪 Concept Checks: `Generators`

🧪 Concept Checks: `Vectorization`

What happens if you try to put a string into an array of integers? Explain `dtype` casting.

What does the `.shape` attribute return? What type of object is it?

Explain the difference between `np.arange()` and `np.linspace()`.

How does NumPy handle missing data (NaN) compared to Python `None`?

Write a one-liner to generate an array of 100 random numbers drawn from a normal distribution (`np.random.randn`).

What is the output of `np.array([1, 2]) + np.array([3, 4, 5])`? Explain what happens.

How do you explicitly force a NumPy array to be floats instead of ints upon creation? (Hint: `dtype=float`).

Explain what `np.zeros_like(arr)` does.

What is the difference between `np.random.rand()` and `np.random.randint()`?

How do you find the total memory consumed by a NumPy array? (Hint: `.nbytes`).

Why is it a bad idea to use `append()` in a loop with NumPy arrays?

How do you flatten a 2D matrix into a 1D array? (`.flatten()` or `.ravel()`).

What happens when you do `arr > 5`? What does it return?

How do you get the transpose of a matrix in NumPy? (Hint: `.T`).

What is the dot product of two vectors, and how do you compute it in NumPy? (`np.dot` or `@`).

How do you find the maximum value in an array? How do you find its index? (`argmax`).

What is the difference between `np.nan` and `np.inf`?