Pattern Recognition in Machine Learning

Computational Thinking & Pattern Recognition

---

This gives us a very important idea:

Machine Learning allows computers to learn patterns from examples rather than requiring humans to write every rule explicitly.

Consider traditional programming.

Traditional approach

We tell the computer:

IF condition A

AND condition B

THEN output X.

The programmer provides the rules.

Machine Learning approach

We provide:

Examples

+

Correct answers

The machine learns patterns that connect the examples to the answers.

Conceptually:

DATA -> PATTERN LEARNING -> MODEL

Then:

NEW DATA -> MODEL –

---

An Example: Ripe and Unripe Bananas

Suppose we want to create an AI system that identifies whether a banana is ripe.

We collect photographs.

Training examples

Image	Colour	Shape	Predicted Class
Image 1	Green	Curved	Unripe
Image 2	Green	Curved	Unripe
Image 3	Yellow	Curved	Ripe
Image 4	Yellow	Curved	Ripe
Image 5	Yellow with spots	Curved	Ripe

The system may discover patterns involving:

• colour
• spots
• texture
• shape

Now we give it a new image.

The system examines its features and predicts:

Ripe

This is pattern recognition in action.

---

Pattern ≠ Rule

This is an important distinction.

A rule is explicitly given.

A pattern may need to be discovered.

Rule

If the traffic light is red, stop.

Someone has given us the rule.

Pattern

Suppose we give a computer thousands of examples of emails labelled:

Spam

and

Not Spam

The computer may discover patterns such as:

• suspicious words
• unusual links
• repeated advertising phrases
• suspicious sender behaviour

The patterns are learned from data.