Machine Learning Algorithms - Interactive Study Guide

🎯 What is Machine Learning?

Definition

Machine Learning (ML) is a part of Artificial Intelligence that enables computers to learn from data and make predictions or decisions without being explicitly programmed for each task.

Think of it this way:

Imagine teaching a child to identify fruits:

🍎 You show them many apples → they learn what apples look like
🍌 You show them many bananas → they learn what bananas look like
🍊 Later, they can identify new fruits they've never seen before!

This is exactly how Machine Learning works! Instead of programming rules, we show the computer examples, and it learns patterns.

How is ML Different from Traditional Programming?

Traditional Programming	Machine Learning
👨‍💻 Programmer writes specific rules	🤖 Computer learns rules from data
📝 Input: Rules + Data → Output	📊 Input: Data + Output → Rules
🔧 Need to update code for new scenarios	📈 Improves automatically with more data
Example: Calculator (fixed operations)	Example: Spam filter (learns from emails)

🔑 Key Takeaways

Machine Learning enables computers to learn from experience
It's used in many applications you interact with daily
Instead of programming rules, we provide data and examples
ML systems improve their performance over time

🔢 Types of Machine Learning

There are three main types of Machine Learning, each used for different purposes. Let's understand them with simple examples!

1. Supervised Learning 👨‍🏫

What is it? Learning with a teacher! The computer is given labeled data (data with answers) to learn from.

Think of it as: Learning math with a textbook that shows both questions AND answers.

📧 Example: Email Spam Detection

Training Phase:

Email 1: "Congratulations! You won $1000!" → SPAM
Email 2: "Meeting tomorrow at 3 PM" → NOT SPAM

After Learning: When a new email arrives, the system can predict if it's spam or not!

Two Types of Supervised Learning:

📈 Regression

Predicts continuous numbers

Examples: Predicting house prices, forecasting temperature.

🏷️ Classification

Predicts categories

Examples: Spam/Not Spam, Dog/Cat.

2. Unsupervised Learning 🔍

What is it? Learning without a teacher! The computer finds patterns in data without being told what to look for.

🛍️ Example: Customer Grouping

ML finds groups like "electronics buyers", "grocery buyers", etc. Useful for targeting and personalization.

3. Reinforcement Learning 🎮

What is it? Learning by trial and error with rewards.

🎮 Example: Teaching AI to Play a Game

AI takes action, receives reward or penalty
Adjusts strategy to maximize reward over time
Used in robots, games, and control systems

📊 Understanding Correlation

What is Correlation?

Correlation measures how two things are related to each other. When one thing changes, does the other thing change too?

Types of Correlation

📈 Positive

Both increase together. Example: Study time ↑ → Grades ↑

📉 Negative

One increases, other decreases. Example: Exercise ↑ → Weight ↓

➡️ None

No relationship. Example: Shoe size ≠ Intelligence

Pearson's Correlation Formula:

r = Σ(x - x̄)(y - ȳ) / √[Σ(x - x̄)² × Σ(y - ȳ)²]

⚠️ Correlation ≠ Causation

Two things may move together due to a third cause (e.g., hot weather → both ice-cream sales and swimming rise).

Interactive Demo: Correlation Calculator

Enter two sets of numbers (comma separated):

X values:

Y values:

Enter values above and click Calculate Correlation.

📈 Regression (Linear)

What is Regression?

Regression predicts a continuous value (a number). Linear regression fits a straight line to data points to predict outputs from inputs.

Key Ideas

Dependent variable (y) — what we want to predict (e.g., marks).
Independent variable (x) — input used to predict (e.g., hours studied).
Line equation: y = m*x + c (m = slope, c = intercept)

Plot the data (scatter plot)

Find the best-fit line (least squares)

Use line to predict new y from x

Interactive Demo: Fit a Line

Use the same input fields in the Correlation demo (they share the canvas). After computing correlation, press Fit Line to draw best-fit line and see slope/intercept.

Fit a line after providing X and Y series in Correlation section.

🏷️ Classification (k-NN)

What is Classification?

Classification predicts categories or labels (e.g., spam or not spam). It's a common supervised learning task.

k-Nearest Neighbour (k-NN) — Simple idea

Store labeled examples (training data).
For a new item, find the k closest examples (neighbors).
Use majority label among neighbors to assign label.

Example: Classifying Fruit

Given fruits with size and color, k-NN uses nearest labelled fruits to decide the label of a new fruit.

When to use k-NN?

Small datasets and simple problems.
When you have clear feature values and distance makes sense.
Be careful: wrong feature scales make distances meaningless (normalize features).

🎯 Clustering (k-Means)

What is Clustering?

Clustering groups similar items together without labels. It is a key unsupervised technique (Teacher Handbook: Unit on Machine Learning algorithms).

k-Means — Simple steps

Choose number of clusters k.
Initialize k centroids (randomly).
Assign each point to nearest centroid.
Update centroids to mean of assigned points.
Repeat until assignments stop changing.

Tips

Try multiple initializations — different starts give different clusters.
k-Means works well for roundish clusters; it struggles with weird-shaped clusters.
Choose k using domain knowledge or methods like elbow method.

✅ Quick Quiz

Question will appear here