What Really Counts?
Choose the right details to help AI learn better.
What you'll learn in this level
- How to tell useful features from useless ones
- Why more features aren't automatically better
- What overfitting means
Pick the Right Features
When we teach computers to recognise patterns, we need to decide which details or 'features' to pay attention to.
For example, if you want to tell apples and oranges apart, you might look at their colour, shape, and texture - these are the 'features' you're using to classify them.
Why the choice matters
- 1
Better Accuracy
With the right features the computer predicts better. Features that do not fit only confuse it.
- 2
Faster Learning
With few features the computer finds patterns faster. It also needs less memory.
- 3
No Memorising
With too many features the computer memorises the training data. Then it misses the real patterns.
- 4
How Doctors Do It
Doctors do not run every possible test. They pick the tests that fit how you feel.
Here you try out different features. Which ones help the computer sort best?
Guess first, then experiment
You give the model as many features as you can. Even useless ones like the shelf number. What happens?