= {1,..., K}. Show Exercise 1.2. Suppose we are in a multi-class classification setting, so that Y that the optimal predictor f : XY minimizing the zero-one loss is still the MAP predictor: f(x) = arg max p(y | x). yY