Editing Supervised learning (section)

==Generative training==

The training methods described above are ''discriminative training'' methods, because they seek to find a function <math>g</math> that discriminates well between the different output values (see [[discriminative model]]). For the special case where <math>f(x,y) = P(x,y)</math> is a [[joint probability distribution]] and the loss function is the negative log likelihood <math>- \sum_i \log P(x_i, y_i),</math> a risk minimization algorithm is said to perform ''generative training'', because <math>f</math> can be regarded as a [[generative model]] that explains how the data were generated. Generative training algorithms are often simpler and more computationally efficient than discriminative training algorithms. In some cases, the solution can be computed in closed form as in [[Naive Bayes classifier|naive Bayes]] and [[linear discriminant analysis]].