Learning logic programs by discovering where not to search

TL;DR

This paper introduces a constraint-driven approach to inductive logic programming that identifies where not to search using background knowledge, significantly reducing learning times and enabling scaling to large domains.

Contribution

It presents a novel method to discover constraints from background knowledge to guide ILP search, improving efficiency and scalability.

Findings

Reduced learning times by up to 97%.

Successfully scaled to domains with millions of facts.

Effective across multiple domains including program synthesis and game playing.

Abstract

The goal of inductive logic programming (ILP) is to search for a hypothesis that generalises training examples and background knowledge (BK). To improve performance, we introduce an approach that, before searching for a hypothesis, first discovers where not to search. We use given BK to discover constraints on hypotheses, such as that a number cannot be both even and odd. We use the constraints to bootstrap a constraint-driven ILP system. Our experiments on multiple domains (including program synthesis and game playing) show that our approach can (i) substantially reduce learning times by up to 97%, and (ii) scale to domains with millions of facts.