What is a characteristic feature of discriminative models in AI?

Discriminative models are focused on modeling the decision boundary between classes rather than generating new data. The characteristic feature that distinguishes these models is their ability to differentiate between classes based on observed features. This means that they analyze existing data points to understand how to categorize or classify new, unseen data.

In contrast to generative models, which aim to learn the distribution of data to generate new instances, discriminative models concentrate on predicting the output variable given the input features by learning the conditional probability distribution of the output given the input. This makes them particularly effective for tasks like classification, where identifying the correct label for a data point based on its features is crucial.

The specific focus on observed features enables discriminative models to efficiently make predictions and perform well in classification tasks, as they learn the characteristics that best separate different classes within the training data.