Learning with distance

Abstract

The two main, competing, paradigms in Artificial Intelligence are the numeric (vector-space) and the symbolic approaches. The debate on which approach is the best for modelling intelligence has been called the ’central debate in AI’. ETS is an inductive learning model that unifies these two, competing, approaches to learning. ETS uses a distance function to define a class and also uses distance to direct the learning process. An ETS algorithm is applied to the Monk’s Problems, a set of problems designed to evaluate the performance of modern learning algorithms - whether numeric and symbolic.