ML 101
M01 · L01
Introduction

What is Machine Learning?

A lamp-lit sorting hall: envelopes travel down a long conveyor between rows of sorting bins and pigeonhole shelves.
A sorting machine sends each envelope to a bin. Machine learning teaches computers to make decisions like this from data, not from rules.

Teaching computers to learn from data — not from rules. From spam filters to self-driving cars, ML is reshaping every industry on the planet.

01 / 13
ML 101
M01 · L01
The Big Idea

The Paradigm Shift

Traditional programming writes rules by hand. Machine learning flips this — give the computer data and answers, and let it discover the rules on its own.

The Flip
Old way: Rules + Data → Answers
ML way: Data + Answers → Rules
02 / 13
ML 101
M01 · L01
Comparison

Traditional vs Machine Learning

Try writing the rules for “is there a face in this photo?” by hand. Every rule you think of has an exception.

Traditional
“Two eyes above a nose” — fails on a side view, a shadow, a hat
ML
Thousands of labelled photos — it learns what a face looks like
03 / 13
ML 101
M01 · L01
1943 – 2012

A Brief History

From the first artificial neuron to the deep learning revolution — nearly 70 years of breakthroughs.

1943 — McCulloch-Pitts artificial neuron model
1957 — Perceptron: first trainable neural network
1997 — Deep Blue defeats Kasparov at chess
2012 — AlexNet wins ImageNet — deep learning explodes
04 / 13
ML 101
M01 · L01
Type 1

Supervised Learning

You provide labeled examples — input-output pairs — and the model learns to map new inputs to the correct outputs.

Example
Show 10,000 photos labeled “cat” or “dog” → model classifies new photos automatically
05 / 13
ML 101
M01 · L01
Type 2

Unsupervised Learning

No labels — the algorithm finds hidden patterns and structure in the data on its own. Think of grouping customers by behavior without telling it the groups.

Applications
Customer segmentation, anomaly detection, data compression, topic modeling
06 / 13
ML 101
M01 · L01
Type 3

Reinforcement Learning

Learn by trial and error. An agent takes actions, receives rewards or penalties, and learns the best strategy over time.

Famous examples
AlphaGo beat the world champion, robots learn to walk, game AI masters Atari
07 / 13
ML 101
M01 · L01
The Process

The ML Pipeline

Building a model is a multi-step process — every step matters.

  • Collect — gather raw data
  • Preprocess — clean, normalize, split
  • Train — fit the model to data
  • Evaluate — test on unseen data
08 / 13
ML 101
M01 · L01
The Math

The Math Behind It

At its core, a model learns weights and a bias that map inputs to a prediction.

Linear Model
\hat{y} = w_1 x_1 + w_2 x_2 + \cdots + w_n x_n + b

x the inputs (say, a house’s size and rooms) · w how much each input counts · b a fixed starting amount, called the bias · ŷ (“y-hat”) the prediction.

09 / 13
ML 101
M01 · L01
The Challenge

Overfitting vs Underfitting

The bias-variance tradeoff is the central challenge of machine learning. Here bias means error from a model too simple for the pattern, and variance error from one so flexible it chases noise — not the b of the last slide.

Underfitting
Too simple — misses patterns
Overfitting
Too complex — memorizes noise
10 / 13
ML 101
M01 · L01
Applications

ML Everywhere

You interact with ML dozens of times a day, often without realizing it.

  • Recommendations — Netflix, YouTube, Amazon
  • Medical diagnosis — cancer detection in scans
  • Fraud detection — real-time transaction monitoring
  • Autonomous vehicles — vision + decision-making
11 / 13
ML 101
Knowledge Check

Check what stuck

Three questions from this lesson. Answer to see why — the explanation appears whether you were right or wrong. Your score is not saved.

Question 1 of 0
Score 0/0

12 / 13
ML 101
Summary
Recap

What you learned

Machine learning is the art of finding patterns in data. Three paradigms, one pipeline, and a constant battle between underfitting and overfitting.

Next Lesson
Supervised Learning →
All lessons Read article
13 / 13