ABSTRACT
One of the principles of Accelerated Discovery is that we can gain insight by concentrating everything we know about an entity into a single comprehensive representation. is representation, in its simplest form, is a vector of feature values, where the features represent all the dierent kinds of things we can know about an entity (its properties, its environment, what it is related to, what it interacts with, etc.). e dimensionality of such a feature vector naturally tends to be extremely high. It is not unusual to see feature vectors with tens of thousands, hundreds of thousands, or even millions of distinct values.
