I have a limited background in this subject, and I have read sentences like: identifying/modelling the underlying process that generates the dataset of a class can be used to recognize future object instances of this class.
My question is: what does the "underlying process" exactly mean? Any illustrative example(s)?
I am dealing with a type of classes that the forces/reasons that have brought them (as categories) into existence are known and understood beforehand. The difficulty comes when objects are to be assigned into one or none of these categories.
In other words, the forces that have generated such classes/categories are known, and some of their characteristics are vaguely and abstractly know (obviously!), but not known concretely enough to the point objects can be easily and unambiguously classified.
I assume that the "underlying process" refers to, in this case, the complex features and feature relationships that characterize/define such classes. Am I correct in this assumption?