Learning logic programs by discovering higher-order abstractions

TL;DR

This paper presents Stevie, a system that automatically refactors logic programs by discovering higher-order abstractions, leading to improved learning performance and transferability across domains.

Contribution

It introduces the higher-order refactoring problem and formulates it as a constraint optimization problem, demonstrating significant performance improvements in logic program learning.

Findings

Predictive accuracies improved by 27%

Learning times reduced by 47%

Discovered abstractions transfer across multiple domains

Abstract

We introduce the higher-order refactoring problem, where the goal is to compress a logic program by discovering higher-order abstractions, such as map, filter, and fold. We implement our approach in Stevie, which formulates the refactoring problem as a constraint optimisation problem. Our experiments on multiple domains, including program synthesis and visual reasoning, show that refactoring can improve the learning performance of an inductive logic programming system, specifically improving predictive accuracies by 27% and reducing learning times by 47%. We also show that Stevie can discover abstractions that transfer to multiple domains.