Goal is directly model & estimate posterior probability
클래스 간 차이에 집중해 학습
We define a generative model as fully specifying. For non-generative Discriminative Model model, we are only interested in a conditional mapping Posterior
직접 모델링하기 때문에, 보통 분류 작업에서 더 효율적이고 정확한 결과
Require positive and negative training data, Can be hard to interpret
Discriminative Learning Methods