Baidu Encyclopedia version
The discriminant model in the machine learning domain is a method of modeling the relationship between unknown data y and known data x. The discriminant model is a method based on probability theory. The input variable x is known, and the discriminant model predicts y by constructing the conditional probability distribution P(y|x).
Discriminant models, also known as conditional models, are a type of model used for statistical classification, especially in supervised machine learning. Discriminant classifiers attempt to model by relying solely on observed data while learning how to classify from given statistics. The methods used in supervised learning can be divided into discriminant models or generated models. Compared with the generated model, the discriminant model has fewer assumptions about the distribution, but it depends largely on the quality of the data.
For example, given a set of dog and rabbit tag images, the discrimination model matches the new unlabeled picture with the most similar tag image and then gives the tag class, dog or rabbit. However, the build will develop a model that should be able to output class tags from the assumptions they make to unmarked images, just like all rabbits have red eyes. Typical discriminative learning methods include logistic regression (LR), support vector machines (SVM)), conditional random field (CRF) (specified on the undirected graph), etc. Typical generation model methods include naive Bayes, Gaussian mixture models, and so on.