Generalisation Through Negation and Predicate Invention

TL;DR

This paper presents NOPI, an ILP system that combines negation and predicate invention to improve generalisation, accuracy, and efficiency in learning logic programs across various domains.

Contribution

It introduces a novel ILP approach integrating negation and predicate invention, enabling learning of more expressive logic programs with better generalisation.

Findings

Improved predictive accuracies across multiple domains

Faster learning times compared to existing methods

Ability to learn stratified negation logic programs

Abstract

The ability to generalise from a small number of examples is a fundamental challenge in machine learning. To tackle this challenge, we introduce an inductive logic programming (ILP) approach that combines negation and predicate invention. Combining these two features allows an ILP system to generalise better by learning rules with universally quantified body-only variables. We implement our idea in NOPI, which can learn normal logic programs with predicate invention, including Datalog programs with stratified negation. Our experimental results on multiple domains show that our approach can improve predictive accuracies and learning times.