General solution using ML

We now apply the method of Section 2.3.5. The general idea is to view the features as parameter estimates. To do this, we need a parametric form $p({\bf x};$   $\theta$$)$, depending on a set of parameters that are equal to or closely related to the feature ${\bf z}$. The PDF and the parameters are problem-specific. Therefore, we need to consider each case separately.