What is b in your prediction function f(x)? You never define or update it in your pseudocode, or did I miss something?

Having a real valued target variable means that for each 2 dimensional input data point you observe a real number instead of the categories red and green. If you want to obtain the value of the target variable of a new input data point via regression, you can approximate it by the mean of target variables belonging to the k nearest input data points around the new input point.