Machine Learning of Regression Models¶
Overview of regression models¶
Regression models differ from generative uncertainty models in that they model the effect of one variable on another variable. In the language of probability theory this involves modelling a conditional probability distribution, rather than a joint probability distribution. In some fields regression models are known as discriminative models, but this is usually when the dependent variable in the probability distribution is discrete, and the problem to be solved involves classification [54]. This thesis is concentrated only on the case of continuous variables. In this chapter, we review different classes of regression model and describe how they can be trained and validated from data. We describe in which circumstances each type of model should be used.