Symmetry Perception by Deep Networks: Inadequacy of Feed-Forward Architectures and Improvements with Recurrent Connections

TL;DR

This paper investigates how deep neural networks perceive symmetry, revealing that feed-forward architectures struggle with this task while recurrent networks succeed, highlighting the importance of recurrence for symmetry perception.

Contribution

The study demonstrates the inadequacy of feed-forward DNNs for symmetry perception and shows recurrent architectures can effectively learn symmetry, suggesting a key role for recurrence.

Findings

Feed-forward DNNs fail to learn symmetry perception.

Recurrent architectures successfully learn symmetry.

Recurrent connections are crucial for symmetry perception.

Abstract

Symmetry is omnipresent in nature and perceived by the visual system of many species, as it facilitates detecting ecologically important classes of objects in our environment. Symmetry perception requires abstraction of long-range spatial dependencies between image regions, and its underlying neural mechanisms remain elusive. In this paper, we evaluate Deep Neural Network (DNN) architectures on the task of learning symmetry perception from examples. We demonstrate that feed-forward DNNs that excel at modelling human performance on object recognition tasks, are unable to acquire a general notion of symmetry. This is the case even when the DNNs are architected to capture long-range spatial dependencies, such as through `dilated' convolutions and the recently introduced `transformers' design. By contrast, we find that recurrent architectures are capable of learning to perceive symmetry by decomposing the long-range spatial dependencies into a sequence of local operations, that are reusable for novel images. These results suggest that recurrent connections likely play an important role in symmetry perception in artificial systems, and possibly, biological ones too.