Linear Regression with Train-Test Split From Scratch | ML Foundation Day 2

Описание: This video is Day 2 of Week 1 (Machine Learning Recap) in a long-term AI Engineer Foundations series.

In this lesson, we implement Linear Regression from scratch using NumPy, but this time with the full machine learning workflow:
Train–test split
Proper gradient descent
Model evaluation using Mean Squared Error
The focus of this video is generalization, not curve fitting.

What you’ll learn in this video:
Why machine learning is about learning rules, not memorizing data
What training data and testing data actually represent
How a train–test split works internally (no libraries, no magic)
How gradient descent updates weights correctly
How predictions are generated after training
How Mean Squared Error evaluates model performance
Why training error alone is meaningless
Every function is written manually.
Every mathematical step is explained.
Every abstraction is removed.

Core idea of this lesson:
If you never test on unseen data, you don’t have a model —
you have false confidence.

This video is for:
Aspiring machine learning and AI engineers
Software engineers revisiting fundamentals properly
Learners who want to think like ML engineers, not library users
Anyone who wants to understand what libraries like scikit-learn actually do

What this video intentionally avoids:
No scikit-learn
No shortcuts
No buzzwords
No “just trust the library” explanations
Frameworks like scikit-learn, PyTorch, and TensorFlow all follow this same structure.
This video shows what’s underneath.