Compositionality Through Language Transmission, using Artificial Neural Networks

TL;DR

This paper explores how an Iterated Learning Model (ILM) applied to artificial neural networks affects compositionality, showing modest improvements and complex relationships between accuracy and structure, especially with high-dimensional inputs.

Contribution

It introduces an ILM-based approach for neural networks and evaluates its impact on compositionality, highlighting differences from traditional grammar-based methods.

Findings

ILM leads to modest compositionality improvements

ILM can cause anti-correlation between accuracy and topologic similarity

ILM enhances compositionality with high-dimensional image inputs

Abstract

We propose an architecture and process for using the Iterated Learning Model ("ILM") for artificial neural networks. We show that ILM does not lead to the same clear compositionality as observed using DCGs, but does lead to a modest improvement in compositionality, as measured by holdout accuracy and topologic similarity. We show that ILM can lead to an anti-correlation between holdout accuracy and topologic rho. We demonstrate that ILM can increase compositionality when using non-symbolic high-dimensional images as input.