Learning from many photos

In the picture

1. 1A square photo card on the belt: one example cat or dog
2. 2The dark computer housing with a lit slot: where the examples are read in
3. 3The hovering square photo: a new picture, never seen before
4. 4The bright outline around the animal: the answer the program chose

Examples

• A spam filter learning from the emails you marked as junk.
• A phone keyboard learning the words you type most often.

Quiz: Learning from many photos (Age 15)

1. 1

   According to the card, how does a machine learning program mainly get better at a task?

   • By being told the correct answer for every new case
   • By training on many examples instead of following only hand-written rules
   • By following only rules that people write by hand
2. 2

   What does the card say about using more varied examples for training?

   • They usually lead to better answers on new pictures
   • They make the program need no labelled photos
   • They only help if every future picture is already stored
3. 3

   Why can training on many examples help a program handle cases nobody wrote rules for?

   • The program learns from the examples rather than only from explicit rules
   • The program already contains a rule for every possible case
   • The program guesses randomly until a person writes a rule

The answers, each with its explanation, are in the app. This lesson has left the free week, so its quiz is part of the subscription.