Limited weights neural networks: Very tight entropy based bounds
Valeriu C. Beiu, Sorin Drăghici · University of North Texas Digital Library (University of North Texas) · 1997
Being given a set of m examples (i.e., data-set) from IR{sup n} belonging to k different classes, the problem is to compute the required number-of-bits (i.e., entropy) for correctly classifying the data-set. Very tight upper and lower bounds for a dichotomy (i.e., k = 2) will be presented, but they are valid for the general case.