Learning, Bayesian Probability, Graphical Models, and Abduction

Abstract

This chapter explores the relationship between learning (induction) and abduction. I take what can be called the Bayesian view, where all uncertainty is reflected in probabilities. In this chapter I argue that, not only can abduction be used for induction, but that most current learning techniques (from statistical learning to neural networks to decision trees to inductive logic programming to unsupervised learning) can be best viewed in terms of abduction.