THINK: a system of classification rules
Fernando Moura-Pires, Adolfo Steiger-Garcao · 2002
A formalism to describe the problem of learning by example is proposed, and an algorithm for rule-based classification is presented. The notions of 'known,' 'dc' (don't care) and 'unknown' rules are introduced, and the differences between them are described. Canonical, elementary and essential rules are explained. Two different processes for finding the elementary rules starting from the canonical rules are demonstrated.>