The whole station crew helps label photos. Raven labels. Rocket double checks.
"Hundreds of labels!" Rocket says, stretching his arms.
"What do you notice?" Raven asks. "The model only knows what our labels taught it."
Nova hovers by the finished stack. "Here is a hint," she says. "Who really taught the sorting model?"
A learning model does not decide what is true by itself.
People collect the examples. People usually add the labels.
So the model learns from people's choices.
Careful people make better training data.
| Week review | True or false? |
|---|---|
| A label is the right answer for an example. | ? |
| Training data is a big set of labeled examples. | ? |
| Machine learning needs only one or two examples. | ? |
| People usually supply training data. | ? |
| A model decides what is true without any data. | ? |
Week 4: the crew built a decision tree. It missed the metal ruler.
Week 5: Rocket tried more rules. They broke on shiny rocks.
Week 6: the crew trained a model on labeled examples. It found the straight edges pattern.
Next week, they test the model on photos it has never seen.
Amazing week, explorer! Tomorrow your family trains a person with examples.