Python for AI & ML - Day 11: Introduction to NumPy

Goal: Master NumPy, the foundation of numerical computing in Python, and understand its critical role in AI/ML for handling large datasets and performing fast mathematical operations.

---

1. What is NumPy?

NumPy (Numerical Python) is a library for efficient array processing and vectorized operations. It is the backbone of nearly every scientific Python library, including Pandas, SciPy, and TensorFlow.

Why NumPy for AI/ML?

• Speed: Optimized C/C++ backend for fast computations.

• Memory Efficiency: Stores data in contiguous memory blocks.

• Vectorization: Avoid loops with array operations (e.g., array * 2).

• Broadcasting: Perform operations on arrays of different shapes.

---

2. Installing NumPy

pip install numpy  

Import the library:

import numpy as np  # Standard alias  

---

3. The NumPy Array (ndarray)

Key Features

• Homogeneous (all elements have the same data type).

• Fixed size (resizing creates a new array).

• Supports vectorized operations (element-wise computations).

Creating Arrays

# From a list  
arr1 = np.array([1, 2, 3])  # 1D array  
arr2 = np.array([[1, 2], [3, 4]])  # 2D array  

# Special arrays  
zeros = np.zeros((3, 3))  # 3x3 matrix of zeros  
ones = np.ones((2, 4))    # 2x4 matrix of ones  
identity = np.eye(3)      # 3x3 identity matrix  
range_arr = np.arange(0, 10, 2)  # [0, 2, 4, 6, 8]  
linspace = np.linspace(0, 1, 5)  # 5 evenly spaced values: [0., 0.25, 0.5, 0.75, 1.]  

---

4. Array Attributes

---

5. Array Operations

Reshaping

arr = np.arange(6).reshape(2, 3)  # [[0, 1, 2], [3, 4, 5]]  

Stacking

a = np.array([1, 2])  
b = np.array([3, 4])  
np.vstack((a, b))  # Vertical stack → [[1, 2], [3, 4]]  
np.hstack((a, b))  # Horizontal stack → [1, 2, 3, 4]  

Mathematical Operations

arr = np.array([1, 2, 3])  
print(arr + 2)   # [3, 4, 5] (element-wise addition)  
print(arr * 3)   # [3, 6, 9]  
print(arr ** 2)  # [1, 4, 9]  

Aggregation Functions

arr = np.array([[1, 2], [3, 4]])  
print(np.sum(arr))     # 10 (sum all elements)  
print(np.mean(arr))    # 2.5 (mean)  
print(np.max(arr, axis=0))  # [3, 4] (max along columns)  

---

6. Indexing and Slicing

Works similarly to Python lists but supports multi-dimensional slicing.

1D Array

arr = np.array([0, 1, 2, 3, 4])  
print(arr[1:4])  # [1, 2, 3]  

2D Array

matrix = np.array([[1, 2, 3], [4, 5, 6], [7, 8, 9]])  
print(matrix[1, 2])     # 6 (second row, third column)  
print(matrix[:, 0])     # [1, 4, 7] (first column)  
print(matrix[1:3, 0:2]) # [[4, 5], [7, 8]] (submatrix)  

---

7. Universal Functions (ufunc)

Precompiled C functions for fast element-wise operations:

arr = np.array([1.5, 2.3, -3.7])  
print(np.abs(arr))    # [1.5, 2.3, 3.7]  
print(np.sqrt(arr))   # Square roots  
print(np.exp(arr))    # Exponentials  
print(np.sin(arr))    # Trigonometric functions  

---

8. Broadcasting

NumPy automatically "broadcasts" smaller arrays to match larger ones for operations:

a = np.array([[1, 2], [3, 4]])  # Shape (2, 2)  
b = np.array([10, 20])           # Shape (2,)  
print(a + b)  # [[11, 22], [13, 24]] (b is broadcast to [[10,20], [10,20]])  

Rules for Broadcasting:

1. Dimensions are compared from right to left.

2. Each dimension must either be equal or one of them must be 1.

---

9. Why NumPy Over Python Lists?

FeatureNumPy ArraysPython ListsSpeedOptimized C codeSlower for large dataMemoryContiguous memory blocksScattered memoryFunctionalityBuilt-in math functionsRequires loopsData TypesHomogeneousHeterogeneous

---

10. Real-World AI/ML Applications

1. Image Processing: Represent images as 3D arrays (height × width × RGB).

2. Dataset Handling: Convert CSV data to arrays for model training.

3. Linear Algebra: Matrix operations for neural networks.

---

Practice Exercise

1. Create a 3x3 matrix of random integers between 0 and 10.

2. Compute the mean of each row.

3. Normalize the matrix by dividing each element by the maximum value.

Solution:

# 1. Create matrix  
matrix = np.random.randint(0, 10, size=(3, 3))  

# 2. Row-wise mean  
row_means = np.mean(matrix, axis=1)  

# 3. Normalize  
normalized = matrix / np.max(matrix)  

---

Key Takeaways

• NumPy arrays are fast, memory-efficient, and enable vectorized operations.

• Indexing/Slicing: Access data in multi-dimensional arrays.

• Broadcasting: Perform operations on arrays of different shapes.

---

What’s Next?

Day 12 introduces Pandas, the go-to library for data manipulation and analysis. You’ll use it to clean, transform, and analyze real-world datasets!

---

By mastering NumPy today, you’ve unlocked the ability to handle numerical computations efficiently—a cornerstone of AI/ML! 🚀