Training a machine with examples
Be able to sort data into two categories and understand that the model learns patterns from examples.
Prerequisites
- APatterns and categoriesrequired
Everyday explanation
How do you teach a computer to recognise cats? You do not write a rule ("cats have pointy ears"). You show examples: a hundred images labelled "cat" and a hundred labelled "not cat".
The computer looks for what the cat images have in common on its own. This is called training the machine.
More and more varied examples → better machine. Only white cats in the examples → the machine thinks black cats are not cats!
Interactive
Be the machine. A colleague puts ten customer cases into two piles without stating the rule: "here" and "there". Look at the piles. What do the cases in the "here" pile have in common?
Now you get a new case. Which pile do you put it in? The colleague says right or wrong. That is training: examples in, guess, correction, better guess.
Mastery means
- Sorts examples into two categories and identifies common traits
- Explains that the machine learns from examples, not from an explicit rule
Sign in to do the exercises and build your mastery up.
Sources
- CS Unplugged (CC BY-SA 4.0) — CC BY-SA 4.0
- Teachable Machine (Google, gratis) — free web service