🧠 Introduction

NumPy (Numerical Python).
It’s the foundation of almost every analytical and ML operation — efficient, powerful, and surprisingly elegant.

Let’s go step-by-step with examples.

⚙️ Quick Setup & Why NumPy

Install it easily:

pip install numpy

Why NumPy?

• Efficient, typed arrays (ndarray) — compact and lightning-fast.

• Vectorized operations — no slow Python loops.

• Broadcasting rules — work with arrays of different shapes easily.

💬 “Why not just use Python lists?”
NumPy arrays store homogeneous types and support vectorized math, making them faster and memory-efficient compared to Python lists.

🔸 Arrays vs Python Lists

import numpy as np

lst = list(range(10))
print(lst)

arr = np.arange(10)
print(arr)

Output:

[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

Key Differences

In short:
👉 NumPy arrays = speed + simplicity.

🧩 Creating Arrays & Matrices

Let’s create a 5×5 matrix with values from 1 to 25:

mat1 = np.arange(1,26).reshape(5,5)
print(mat1)

Output:

[[ 1  2  3  4  5]
 [ 6  7  8  9 10]
 [11 12 13 14 15]
 [16 17 18 19 20]
 [21 22 23 24 25]]

Quick Concept: arange vs linspace

• np.arange(start, stop) → steps by 1 (or given step), stop excluded.

• np.linspace(start, stop, num) → divides range into num equal intervals (stop included).

Tip: Mention this difference if asked in an interview — it shows clarity and precision. 😉

🎯 Indexing & Slicing

Extract the middle 3×3 matrix:

mat2 = mat1[1:4, 1:4]
print(mat2)

Result:

[[ 7  8  9]
 [12 13 14]
 [17 18 19]]

Concept:

• Python uses zero-based indexing.

• a:b means start at a, end before b.

So, mat1[1:4, 1:4] selects rows 1–3 and columns 1–3.

💡 Bonus:
To pick the center element, use:

mat1[2, 2]  # returns 13

📊 Aggregations — Sum & Mean

Sum of Each Row

row_sum = np.sum(mat1, axis=1)
print("Row wise sum:", row_sum)

Output:

Row wise sum: [ 15  40  65  90 115]

Mean of Each Column

col_mean = np.mean(mat1, axis=0)
print("Column wise mean:", col_mean)

Output:

Column wise mean: [11. 12. 13. 14. 15.]

Understanding axis

Note:
If you just call mat1.sum() (without axis), it sums all elements and returns a scalar.

🔍 Boolean Indexing — Conditional Selection

Let’s find values greater than 20:

high_values = mat1[mat1 > 20]
print("High values above 20:", high_values)

Output:

High values above 20: [21 22 23 24 25]

How it works

• mat1 > 20 → creates a boolean mask (True/False values).

• mat1[mask] → returns only elements where mask is True.

Example mask:

[[False False ...]
 [False False ...]
 [False False ...]
 [False False ...]
 [ True  True  True  True  True]]

💬 In simple words:
This mask works like a filter — wherever the condition is True, those values get selected.
Thoda sa logic samajh lo, aur tum boolean indexing ke master ban jaoge 😄

💬 Common Questions

Q1: What is broadcasting?
A: It’s NumPy’s way of doing arithmetic on arrays of different shapes by automatically expanding them along missing dimensions.

Q2: How to reshape without copying data?
A: arr.reshape() returns a view if possible; use arr.copy() for a deep copy.

Q3: How to combine arrays?
A: Use np.concatenate, np.vstack, or np.hstack.

Q4: How to get unique elements?
A: np.unique(arr)

Q5: How to handle NaN values?
A: Use np.nanmean, np.nansum, or filter them before aggregation.

🧭 Best Practices

✅ Use .shape to verify dimensions
✅ Check dtype regularly
✅ Avoid Python loops — prefer vectorized ops
✅ Be careful with views vs copies
✅ Print small samples while debugging

🏁 Final Words

NumPy is not just a library — it’s the language of data in Python.
Har din thoda practice, aur jaldi hi aap bolenge — “NumPy? Easy yaar!” 😎