Summary

In this chapter, we learned how to extend the simple linear regression model to deal with categorical predicted data and how to perform Bayesian classification using either logistic regression when we have two classes or softmax regression for more than two classes. We learned what an inverse link function is and how it is used to build Generalized Linear Models (GLM), which extends the range of problems that can be solved by linear models. We also learned about some precautions we have to take, for example, when dealing with correlated variables, perfectly separable classes or unbalanced classes. While we focused on discriminative models for classification, we also learned about generative models and some of the main differences between both types of models.