In-class Activity: Dogs or Cats

The following activity practices K Nearest Neighbors.

[Team] Activity: Dogs or Cats

Here we use features Body Size and Food Intake to classify cats and dogs. The training data are shown as cat or dog images, and there are three test data to be labeled: green, red, and blue points.

Consider K Nearest Neighbors with \(K = 1, 2, 3, 4\), and discuss the following questions.

Discussion

For which test data point, the classification result is uncertain or unstable?

Which data point can be labelled as one particular category with 100% certainty, for all different \(K\)s?

For the blue point, what happened when \(K\) increases? Do we change the prediction? What about uncertainty?

If you would like to make a classification rule or set the decision boundary to predict cats or dogs, what would the decision boundary look like? Why?