Foundation for the New Algorithm Learning Pseudo-Independent Models

Abstract

A type of problem domains known as pseudo-independent (PI) models poses difficulty for common learning methods, which are based on the single-link lookahead search. To learn this type of domain models, a method called the multiple-link lookahead search is needed. An improved result can be obtained by incorporating model complexity into a scoring metric to explicitly trade off model accuracy for complexity and vice versa during selection of the best model among candidates at each learning step. Previous studies found the complexity formulae for full PI models (the simplest type of PI models) and for atomic PI models (PI models without submodels). This study presents the complexity formula for non-atomic PI models, which are more complex than full or atomic PI models, yet more general. Together with the previous results, this study completes the major theoretical work for the new learning algorithm that combines complexity and accuracy.