# GLMs

Let's start our analysis of regression models by defining the context we're working with. A regression is a model that associates an input vector, , with one or more continuous dependent variables (for simplicity, we're going to refer to single outputs), . In a general scenario, there's no explicit dependence on time, even if regression models are often employed to model time series. The main difference is that, in the latter, the order of the data points cannot be changed, because there are often inter-dependencies. On the other hand, a generic regression can be used to model time-independent phenomena, and, in the context of GLMs, we're initially assuming that we work with stateless associations where the output value depends only on the input vector. In such cases, it's also possible to shuffle the dataset without changing the final result (of course, this is not true if the output at time *t* depends, for example, on *y*_{t-1}, which is a function of...