Compositional generalization through meta sequence-to-sequence learning

TL;DR

This paper introduces a meta sequence-to-sequence learning approach with memory-augmented neural networks that enhances compositional generalization, enabling models to understand and apply new concepts compositionally, similar to human learning.

Contribution

It presents a novel meta-learning framework for seq2seq models with memory augmentation that improves compositional generalization capabilities beyond traditional neural networks.

Findings

Meta seq2seq learning solves key SCAN compositionality tests.

Models can learn to apply implicit rules to variables.

Approach outperforms standard seq2seq models on compositional tasks.

Abstract

People can learn a new concept and use it compositionally, understanding how to "blicket twice" after learning how to "blicket." In contrast, powerful sequence-to-sequence (seq2seq) neural networks fail such tests of compositionality, especially when composing new concepts together with existing concepts. In this paper, I show how memory-augmented neural networks can be trained to generalize compositionally through meta seq2seq learning. In this approach, models train on a series of seq2seq problems to acquire the compositional skills needed to solve new seq2seq problems. Meta se2seq learning solves several of the SCAN tests for compositional learning and can learn to apply implicit rules to variables.