A discriminative model predicts labels or outputs from observed features by learning the conditional relationship between inputs and targets. Most supervised classification and regression models are discriminative, in contrast to generative models that learn how data itself is produced.
A model that predicts labels from a set of one or more features. More formally, discriminative models define the conditional probability of an output given the features and weights; that is: For example, a model that predicts whether an email is spam from features and weights is a discriminative model. The vast majority of supervised learning models, including classification and regression models, are discriminative models. Contrast with generative model.