Discriminative model Machine Learning
noun phrase
Definition: A machine-learning model that directly learns a mapping from inputs to outputs or models the conditional distribution P(y∣x), rather than modelling the full joint distribution of data and labels. In contemporary ML usage, discriminative models are contrasted with generative models because they focus on prediction or decision boundaries rather than data generation [Kasa 2025].
Example in context: “… a semantic consistency scorer is a discriminative model that computes the probability that two samples belong to the same class.” [Xie and Li 2024]
Synonyms: conditional model (near-equivalent in some ML contexts); discriminative classifier (narrower)
Related terms: generative model; discriminative classifier; conditional probability model; decision boundary