Learning Compositional Rules via Neural Program Synthesis

TL;DR

This paper introduces a neuro-symbolic model that learns explicit rule systems from few examples, outperforming neural meta-learning in tasks requiring compositional generalization across language and instruction domains.

Contribution

The work presents a novel neural program synthesis approach for learning explicit rule systems, enhancing compositional generalization in neural models.

Findings

Outperforms neural meta-learning in multiple domains

Effective in language translation and instruction learning

Generalizes systematically from limited data

Abstract

Many aspects of human reasoning, including language, require learning rules from very little data. Humans can do this, often learning systematic rules from very few examples, and combining these rules to form compositional rule-based systems. Current neural architectures, on the other hand, often fail to generalize in a compositional manner, especially when evaluated in ways that vary systematically from training. In this work, we present a neuro-symbolic model which learns entire rule systems from a small set of examples. Instead of directly predicting outputs from inputs, we train our model to induce the explicit system of rules governing a set of previously seen examples, drawing upon techniques from the neural program synthesis literature. Our rule-synthesis approach outperforms neural meta-learning techniques in three domains: an artificial instruction-learning domain used to evaluate human learning, the SCAN challenge datasets, and learning rule-based translations of number words into integers for a wide range of human languages.