An Explicitly Relational Neural Network Architecture

TL;DR

This paper introduces a new neural network architecture designed to learn explicit relational structures from raw pixel data, improving visual relational reasoning capabilities and generalization on unseen tasks.

Contribution

The paper proposes a novel end-to-end neural network architecture that explicitly models relational structures, bridging deep learning and symbolic AI.

Findings

Pre-trained on relational tasks, the architecture learns reusable representations.

The architecture outperforms baseline models on unseen tasks.

Visualizations provide insights into the model's functioning.

Abstract

With a view to bridging the gap between deep learning and symbolic AI, we present a novel end-to-end neural network architecture that learns to form propositional representations with an explicitly relational structure from raw pixel data. In order to evaluate and analyse the architecture, we introduce a family of simple visual relational reasoning tasks of varying complexity. We show that the proposed architecture, when pre-trained on a curriculum of such tasks, learns to generate reusable representations that better facilitate subsequent learning on previously unseen tasks when compared to a number of baseline architectures. The workings of a successfully trained model are visualised to shed some light on how the architecture functions.