Track quiz
Before the questions, here is each idea in one line.
| Idea | The line to keep |
|---|---|
| The closest one | Measure the gap to every priced crate. The nearest crate's price is the guess. |
| The whole board | Every square copies its nearest crate. A tie gets no price. Move a crate and the map changes. |
| No labels at all | With no prices, closeness can still make groups. Repeat until nobody changes group. |
| Test on new crates | Test on crates the program never saw. Accuracy is right answers out of all answers. |
| One-sided examples | Examples from one place teach one place. Add examples from the missing places. |
A new crate is 2 across, 1 up from a priced crate. What is the gap number?
Crate P has gap number 5. Crate Q has gap number 10. Whose price does the program copy?
Two nearest crates are equally near. One says full price and one says cut price. What does the program say?
Eight crates sit on a board. Which squares get a price on the map?
You move one crate. What happens to the map?
The crates come with the seller's price on them. What is learning from them called?
The prices are rubbed off ten crates. What can a program still do?
Group A holds 3 full price and 2 cut price crates. What does that show?
When does the grouping stop?
A board guesses 9 of 12 kept-back crates right. What is its accuracy?
Why are ten crates kept back and never put on the board?
A board is right about all four of its own crates. What does that show about new crates?
All six examples came from the early bus, so all are fresh. Which crates does the board get wrong most?
Three more bus crates are added. What happens to the score on older crates?
What fixes a board that has examples from only one place?
Ask two things of any program. Which examples did it learn from? Which new examples was it tested on?
Next: Digital Safety & Citizenship III: What a Machine Hands You. A machine gives you an answer. This track showed what its answer is built from.