In discriminative learning we model \(P(y\mid x)\). In generative learning, what do we typically model? Hint: In generative approaches we specify how data are generated given the class and include the class prior. Using Bayes' rule then yields \(P(y\mid x)\).