Inverting the Resolution Principle
Stephen Muggleton · 1991
Abstract In this paper we describe the current status of an ongoing research project investigating a novel form of Machine Learning in which the learner’s vocabulary is enriched by the machine suggesting useful new descriptive terms for the user to accept or reject. An algorithm called Duce has been shown to be effective along these lines in developing and extending propositional theories within a chess endgame domain and a diagnostic domain of neuro-psychology. By showing that Duce’s trans formational operators are based on reversing the steps of a resolution proof we show that Duce’s learning method is sufficient for learning any propositional theory.