Learning logic programs by explaining their failures

TL;DR

This paper introduces failure explanation techniques for inductive logic programming, enabling more precise hypothesis refinement by analyzing failures at the sub-program level, which significantly improves learning efficiency.

Contribution

It presents a novel failure explanation algorithm based on SLD-tree analysis and integrates it into the Popper ILP system to enhance hypothesis space pruning.

Findings

Failure explanation reduces hypothesis space exploration

Learning times are significantly decreased

Fine-grained failure analysis improves hypothesis refinement

Abstract

Scientists form hypotheses and experimentally test them. If a hypothesis fails (is refuted), scientists try to explain the failure to eliminate other hypotheses. The more precise the failure analysis the more hypotheses can be eliminated. Thus inspired, we introduce failure explanation techniques for inductive logic programming. Given a hypothesis represented as a logic program, we test it on examples. If a hypothesis fails, we explain the failure in terms of failing sub-programs. In case a positive example fails, we identify failing sub-programs at the granularity of literals. We introduce a failure explanation algorithm based on analysing branches of SLD-trees. We integrate a meta-interpreter based implementation of this algorithm with the test-stage of the Popper ILP system. We show that fine-grained failure analysis allows for learning fine-grained constraints on the hypothesis space. Our experimental results show that explaining failures can drastically reduce hypothesis space exploration and learning times.