Predicate Invention by Learning From Failures

TL;DR

This paper introduces POPPI, an ILP system that formulates predicate invention as an answer set programming problem, significantly improving learning performance and outperforming existing systems.

Contribution

The paper presents POPPI, a novel ILP system that effectively incorporates predicate invention by leveraging answer set programming, addressing a longstanding challenge in ILP.

Findings

Predicate invention can significantly enhance ILP performance.

POPPI outperforms existing ILP systems in experiments.

Predicate invention is computationally feasible when useful, and not costly when unnecessary.

Abstract

Discovering novel high-level concepts is one of the most important steps needed for human-level AI. In inductive logic programming (ILP), discovering novel high-level concepts is known as predicate invention (PI). Although seen as crucial since the founding of ILP, PI is notoriously difficult and most ILP systems do not support it. In this paper, we introduce POPPI, an ILP system that formulates the PI problem as an answer set programming problem. Our experiments show that (i) PI can drastically improve learning performance when useful, (ii) PI is not too costly when unnecessary, and (iii) POPPI can substantially outperform existing ILP systems.