Arguments for the unsuitability of convolutional neural networks for non-local tasks

DOI10.1016/J.NEUNET.2021.05.001zbMATH Open1521.68158arXiv2102.11944OpenAlexW3160733000WikidataQ125009793 ScholiaQ125009793MaRDI QIDQ6079068

Sebastian Stabinger, Antonio Rodríguez-Sánchez, David Peer

Publication date: 28 September 2023

Published in: Neural Networks (Search for Journal in Brave)

Abstract: Convolutional neural networks have established themselves over the past years as the state of the art method for image classification, and for many datasets, they even surpass humans in categorizing images. Unfortunately, the same architectures perform much worse when they have to compare parts of an image to each other to correctly classify this image. Until now, no well-formed theoretical argument has been presented to explain this deficiency. In this paper, we will argue that convolutional layers are of little use for such problems, since comparison tasks are global by nature, but convolutional layers are local by design. We will use this insight to reformulate a comparison task into a sorting task and use findings on sorting networks to propose a lower bound for the number of parameters a neural network needs to solve comparison tasks in a generalizable way. We will use this lower bound to argue that attention, as well as iterative/recurrent processing, is needed to prevent a combinatorial explosion.

Full work available at URL: https://arxiv.org/abs/2102.11944

zbMATH Keywords

attention locality convolutional neural networks sorting networks relational reasoning

Mathematics Subject Classification ID

Classification and discrimination; cluster analysis (statistical aspects) (62H30) Learning and adaptive systems in artificial intelligence (68T05) Computing methodologies for image processing (68U10)

Cites Work

Title not available (Why is that?)

This page was built for publication: Arguments for the unsuitability of convolutional neural networks for non-local tasks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6079068)