Think Together · Start with a Short Story
The girl and the robot are preparing a cat-and-dog classification experiment. Before training, they check whether the photos include different poses and backgrounds.
Lay Out the Process
- CollectVaried pictures of cats and dogs
- LabelConfirm the correct category
- TrainAdjust based on the error
- TestUse pictures it has not seen
Teaching a classifier means defining the task, preparing suitable examples, training, and evaluating it independently.
If labels are wrong, the model gets incorrect feedback. If all examples look alike, it may only handle that one situation.
During testing, you can also group results by different backgrounds and poses to see where the model is likely to make mistakes.
Now You Try
The overall accuracy is high, but black cats are often classified incorrectly. Can we ignore that?
Ready? Open this to see the explanation
No. One overall number can hide problems with a particular group. Check whether the data covers that group and examine those errors.
One Idea to Take Away
Teaching a classifier means defining the task, preparing suitable examples, training, and evaluating it independently.
Try covering the diagram above and explaining each step to someone in your family. If one step is hard to explain, return to that card and look again.