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

Levels of Representation in a Deep Learning Model of Categorization

Olivia Guest, Bradley C. Love
doi: https://doi.org/10.1101/626374
Olivia Guest
1Department of Experimental Psychology, University College London
  • Find this author on Google Scholar
  • Find this author on PubMed
  • Search for this author on this site
Bradley C. Love
2Department of Experimental Psychology, University College London, and The Alan Turing Institute
  • 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 (DCNNs) rival humans in object recognition. The layers (or levels of representation) in DCNNs have been successfully aligned with processing stages along the ventral stream for visual processing. Here, we propose a model of concept learning that uses visual representations from these networks to build memory representations of novel categories, which may rely on the medial temporal lobe (MTL) and medial prefrontal cortex (mPFC). Our approach opens up two possibilities: a) formal investigations can involve photographic stimuli as opposed to stimuli handcrafted and coded by the experimenter; b) model comparison can determine which level of representation within a DCNN a learner is using during categorization decisions. Pursuing the latter point, DCNNs suggest that the shape bias in children relies on representations at more advanced network layers whereas a learner that relied on lower network layers would display a color bias. These results confirm the role of natural statistics in the shape bias (i.e., shape is predictive of category membership) while highlighting that the type of statistics matter, i.e., those from lower or higher levels of representation. We use the same approach to provide evidence that pigeons performing seemingly sophisticated categorization of complex imagery may in fact be relying on representations that are very low-level (i.e., retinotopic). Although complex features, such as shape, relatively predominate at more advanced network layers, even simple features, such as spatial frequency and orientation, are better represented at the more advanced layers, contrary to a standard hierarchical view.

Footnotes

  • This work was supported by NIH (Grant 1P01HD080679), and a Wellcome Trust Investigator Award (Grant WT106931MA) to BCL, as well as The Alan Turing Institute under the EPSRC grant EP/N510129/1. Some of this work was originally reported at the 39th Annual Meeting of the Cognitive Science Society in 2017.

  • The authors declare that they have no competing interests. The authors would like to thank the members of the Love lab at UCL for their useful input when preparing the manuscript. The code used to run these experiments is available here: https://osf.io/jxavn/.

  • https://osf.io/jxavn/

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 May 22, 2019.
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.
Levels of Representation in a Deep Learning Model of Categorization
(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
Levels of Representation in a Deep Learning Model of Categorization
Olivia Guest, Bradley C. Love
bioRxiv 626374; doi: https://doi.org/10.1101/626374
Digg logo Reddit logo Twitter logo Facebook logo Google logo LinkedIn logo Mendeley logo
Citation Tools
Levels of Representation in a Deep Learning Model of Categorization
Olivia Guest, Bradley C. Love
bioRxiv 626374; doi: https://doi.org/10.1101/626374

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 (3686)
  • Biochemistry (7767)
  • Bioengineering (5666)
  • Bioinformatics (21237)
  • Biophysics (10553)
  • Cancer Biology (8159)
  • Cell Biology (11904)
  • Clinical Trials (138)
  • Developmental Biology (6737)
  • Ecology (10388)
  • Epidemiology (2065)
  • Evolutionary Biology (13838)
  • Genetics (9694)
  • Genomics (13054)
  • Immunology (8121)
  • Microbiology (19936)
  • Molecular Biology (7825)
  • Neuroscience (42959)
  • Paleontology (318)
  • Pathology (1276)
  • Pharmacology and Toxicology (2256)
  • Physiology (3350)
  • Plant Biology (7207)
  • Scientific Communication and Education (1309)
  • Synthetic Biology (1998)
  • Systems Biology (5528)
  • Zoology (1126)