numpy-for-ml

Project Overview

numpy-for-ml is a professional learning project designed to strip away the complexity of high-level Machine Learning libraries and focus on the fundamental building block of Data Science in Python: NumPy.

This repository contains clear, runnable, and heavily commented examples demonstrating how core ML concepts (like normalization, distance metrics, and matrix multiplication) are implemented using pure NumPy.

Topics Covered

The project is organized into four key modules:

1. Basics

   • array_creation.py: Understanding scalars, vectors, matrices, and data types.
   • indexing_slicing.py: Extracting data, slicing rows/columns, and boolean masking.

2. Operations

   • arithmetic_ops.py: Element-wise comparisons and the power of vectorized computation vs. loops.
   • broadcasting.py: How NumPy handles operations between arrays of different shapes.

3. Linear Algebra

   • dot_product.py: The engine of neural networks (weighted sums).
   • matrix_multiplication.py: Propagating inputs through layers using np.dot and @.

4. ML Foundations

   • normalization.py: Implementing Feature Scaling (Z-score normalization) from scratch.
   • distance_metrics.py: Calculating Euclidean distance for algorithms like K-Nearest Neighbors (KNN).

Why NumPy is Important for ML

• Performance: NumPy uses optimized C-level primitives, making it orders of magnitude faster than Python lists.
• Foundation: Libraries like TensorFlow, PyTorch, Scikit-Learn, and Pandas are all built on top of NumPy.
• Vectorization: It allows for mathematical operations on entire datasets without writing slow loops, which is essential for training models on massive data.

How to Run

Ensure you have Python and NumPy installed.

pip install numpy

Navigate to the project directory and run any script:

# Example: Run the broadcasting demo
python operations/broadcasting.py

# Example: Run the normalization demo
python ml_foundations/normalization.py

Author

Prem Patel