Mission control is buzzing. Rocket has stacked forty snapshot cards beside the telescope screen, each one turned into a row of numbers.
"We have the feature table," Rocket says. "Now Sky Sorter just has to know what every row means."
Raven frowns at the stack. "A row of numbers is not the sky. It only keeps the parts we chose to measure."
Nova floats over the table and projects one row onto the wall: brightness 230, size 5, streak no.
"Every row is a record," she says. "What can you work out from this one, and what is missing?"
Rocket and Raven lean in to look.
Last week the crew turned each snapshot into a feature table.
That table is a representation: a way of storing what we know so a computer can use it.
Computers build representations out of data structures, such as lists, tables and trees.
Each row of the Sky Sorter table is a record. Each column is one feature.
| Card | Brightness (0 to 255) | Size (pixels wide) | Streak? |
|---|---|---|---|
| A | 120 | 2 | no |
| B | 230 | 5 | no |
| C | 90 | 12 | no |
| D | 180 | 1 | yes |
| E | 210 | 3 | no |
| F | 150 | 9 | no |
These are invented practice cards for our story. We will use them all week.
A representation is useful because it is simple.
But it also leaves things out. The table does not store color, shape or the time of the snapshot.
If an important clue is missing from the table, no algorithm can use it.
Choosing what goes into a representation is a design decision people make.
| Statement | True or false? |
|---|---|
| Each row in the Sky Sorter table describes one snapshot. | ? |
| A table, a list and a tree are all data structures. | ? |
| The table stores everything about the real sky. | ? |
| People choose which features go into the table. | ? |
Strong start! Tomorrow we use reasoning to get new answers from these records.