Utility of model-order reduction methods
An all-purpose ideal method for model-order reduction does not exist. Indeed, the context in which model reduction is applied will help determine the best method. The proposed framework allows a natural categorisation of methods based on their utility (illustrated in Table 1).
The model-order reduction techniques discussed in this article have been grouped into parametric and nonparametric methods. Parametric methods are often useful in settings where repeated use of the model particularly in which the parameter values have mechanistic meaning, e.g. in estimation or simulation. In contrast non-parametric methods are valuable for any setting where repeated simulation is the goal. Since non-parametric models do not retain mechanistic structure their use in estimation may be limited.