Skip to main content
bioRxiv
  • Home
  • About
  • Submit
  • ALERTS / RSS
Advanced Search
New Results

Biological convolutions improve DNN robustness to noise and generalisation

View ORCID ProfileBenjamin D. Evans, Gaurav Malhotra, Jeffrey S. Bowers
doi: https://doi.org/10.1101/2021.02.18.431827
Benjamin D. Evans
School of Psychological Science, University of Bristol, 12a Priory Road, Bristol BS8 1TU, UK
  • Find this author on Google Scholar
  • Find this author on PubMed
  • Search for this author on this site
  • ORCID record for Benjamin D. Evans
  • For correspondence: benjamin.evans@bristol.ac.uk
Gaurav Malhotra
School of Psychological Science, University of Bristol, 12a Priory Road, Bristol BS8 1TU, UK
  • Find this author on Google Scholar
  • Find this author on PubMed
  • Search for this author on this site
Jeffrey S. Bowers
School of Psychological Science, University of Bristol, 12a Priory Road, Bristol BS8 1TU, UK
  • Find this author on Google Scholar
  • Find this author on PubMed
  • Search for this author on this site
  • Abstract
  • Full Text
  • Info/History
  • Metrics
  • Data/Code
  • Preview PDF
Loading

Abstract

Deep Convolutional Neural Networks (DNNs) have achieved superhuman accuracy on standard image classification benchmarks. Their success has reignited significant interest in their use as models of the primate visual system, bolstered by claims of their architectural and representational similarities. However, closer scrutiny of these models suggests that they rely on various forms of shortcut learning to achieve their impressive performance, such as using texture rather than shape information. Such superficial solutions to image recognition have been shown to make DNNs brittle in the face of more challenging tests such as noise-perturbed or out-of-domain images, casting doubt on their similarity to their biological counterparts. In the present work, we demonstrate that adding fixed biological filter banks, in particular banks of Gabor filters, helps to constrain the networks to avoid reliance on shortcuts, making them develop more structured internal representations and more tolerant to noise. Importantly, they also gained around 20 35% improved accuracy when generalising to our novel out-of-domain test image sets over standard end-to-end trained architectures. We take these findings to suggest that these properties of the primate visual system should be incorporated into DNNs to make them more able to cope with real-world vision and better capture some of the more impressive aspects of human visual perception such as generalisation.

Competing Interest Statement

The authors have declared no competing interest.

Footnotes

  • Results for VGG-19 were replaced with ResNet50 (Revised Figs 5, 6 & 7); Figures 1&11 were added to illustrate the convolutional kernels; The previous Figure 6 was removed; New results were added in Figures 9, 14--18, examining tolerance to perturbations in with the generalisation test images; Associated minor revisions to the text were made, including revising Equation 5, adding new references and extending the discussion.

  • https://github.com/bdevans/BioNet

Copyright 
The copyright holder for this preprint is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY 4.0 International license.
Back to top
PreviousNext
Posted September 09, 2021.
Download PDF
Data/Code
Email

Thank you for your interest in spreading the word about bioRxiv.

NOTE: Your email address is requested solely to identify you as the sender of this article.

Enter multiple addresses on separate lines or separate them with commas.
Biological convolutions improve DNN robustness to noise and generalisation
(Your Name) has forwarded a page to you from bioRxiv
(Your Name) thought you would like to see this page from the bioRxiv website.
CAPTCHA
This question is for testing whether or not you are a human visitor and to prevent automated spam submissions.
Share
Biological convolutions improve DNN robustness to noise and generalisation
Benjamin D. Evans, Gaurav Malhotra, Jeffrey S. Bowers
bioRxiv 2021.02.18.431827; doi: https://doi.org/10.1101/2021.02.18.431827
Reddit logo Twitter logo Facebook logo LinkedIn logo Mendeley logo
Citation Tools
Biological convolutions improve DNN robustness to noise and generalisation
Benjamin D. Evans, Gaurav Malhotra, Jeffrey S. Bowers
bioRxiv 2021.02.18.431827; doi: https://doi.org/10.1101/2021.02.18.431827

Citation Manager Formats

  • BibTeX
  • Bookends
  • EasyBib
  • EndNote (tagged)
  • EndNote 8 (xml)
  • Medlars
  • Mendeley
  • Papers
  • RefWorks Tagged
  • Ref Manager
  • RIS
  • Zotero
  • Tweet Widget
  • Facebook Like
  • Google Plus One

Subject Area

  • Neuroscience
Subject Areas
All Articles
  • Animal Behavior and Cognition (4229)
  • Biochemistry (9118)
  • Bioengineering (6759)
  • Bioinformatics (23956)
  • Biophysics (12107)
  • Cancer Biology (9505)
  • Cell Biology (13746)
  • Clinical Trials (138)
  • Developmental Biology (7618)
  • Ecology (11669)
  • Epidemiology (2066)
  • Evolutionary Biology (15482)
  • Genetics (10622)
  • Genomics (14302)
  • Immunology (9472)
  • Microbiology (22810)
  • Molecular Biology (9083)
  • Neuroscience (48903)
  • Paleontology (355)
  • Pathology (1479)
  • Pharmacology and Toxicology (2566)
  • Physiology (3829)
  • Plant Biology (8320)
  • Scientific Communication and Education (1468)
  • Synthetic Biology (2294)
  • Systems Biology (6176)
  • Zoology (1297)