Abstract

Color is a prime example of categorical perception, yet it is unclear why and how color categories emerge. On the one hand, prelinguistic infants and several animals treat color categorically. On the other hand, recent modeling endeavors have successfully utilized communicative concepts as the driving force for color categories. Rather than modeling categories directly, we investigate the potential emergence of color categories as a result of acquiring visual skills. Specifically, we asked whether color is represented categorically in a convolutional neural network (CNN) trained to recognize objects in natural images. We systematically trained new output layers to the CNN for a color classification task and, probing novel colors, found borders that are largely invariant to the training colors. The border locations were confirmed using an evolutionary algorithm that relies on the principle of categorical perception. A psychophysical experiment on human observers, analogous to our primary CNN experiment, shows that the borders agree to a large degree with human category boundaries. These results provide evidence that the development of basic visual skills can contribute to the emergence of a categorical representation of color.

Data availability

The main analyses were computational and performed on ResNets from the models module of the torchvision package for python (see https://pytorch.org/vision/). Only Figure 4 is based on human data. Human data and source code for running the analysis and generating figures can be found at: https://github.com/vriesdejelmer/colorCategories/The code for the ipad experiment is available at:https://github.com/vriesdejelmer/ColorCoder/

Article and author information

Author details

  1. Jelmer P de Vries

    Experimental Psychology, University of Giessen, Giessen, Germany
    Competing interests
    The authors declare that no competing interests exist.
  2. Arash Akbarinia

    Experimental Psychology, University of Giessen, Giessen, Germany
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4249-231X
  3. Alban Flachot

    Experimental Psychology, University of Giessen, Giessen, Germany
    Competing interests
    The authors declare that no competing interests exist.
  4. Karl R Gegenfurtner

    Experimental Psychology, University of Giessen, Giessen, Germany
    For correspondence
    gegenfurtner@uni-giessen.de
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5390-0684

Funding

Deutsche Forschungsgemeinschaft (222641018 SFB TRR 135)

  • Jelmer P de Vries
  • Arash Akbarinia
  • Alban Flachot
  • Karl R Gegenfurtner

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Ethics

Human subjects: Informed consent was obtained from all observers prior to the experiment. All procedures were approved by the local ethics committee at Giessen University (LEK 2021-0033).

Copyright

© 2022, de Vries et al.

This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,918
    views
  • 247
    downloads
  • 13
    citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Jelmer P de Vries
  2. Arash Akbarinia
  3. Alban Flachot
  4. Karl R Gegenfurtner
(2022)
Emergent color categorization in a neural network trained for object recognition
eLife 11:e76472.
https://doi.org/10.7554/eLife.76472

Share this article

https://doi.org/10.7554/eLife.76472

Further reading

    1. Biochemistry and Chemical Biology
    2. Computational and Systems Biology
    A Sofia F Oliveira, Fiona L Kearns ... Adrian J Mulholland
    Short Report

    The spike protein is essential to the SARS-CoV-2 virus life cycle, facilitating virus entry and mediating viral-host membrane fusion. The spike contains a fatty acid (FA) binding site between every two neighbouring receptor-binding domains. This site is coupled to key regions in the protein, but the impact of glycans on these allosteric effects has not been investigated. Using dynamical nonequilibrium molecular dynamics (D-NEMD) simulations, we explore the allosteric effects of the FA site in the fully glycosylated spike of the SARS-CoV-2 ancestral variant. Our results identify the allosteric networks connecting the FA site to functionally important regions in the protein, including the receptor-binding motif, an antigenic supersite in the N-terminal domain, the fusion peptide region, and another allosteric site known to bind heme and biliverdin. The networks identified here highlight the complexity of the allosteric modulation in this protein and reveal a striking and unexpected link between different allosteric sites. Comparison of the FA site connections from D-NEMD in the glycosylated and non-glycosylated spike revealed that glycans do not qualitatively change the internal allosteric pathways but can facilitate the transmission of the structural changes within and between subunits.

    1. Computational and Systems Biology
    George N Bendzunas, Dominic P Byrne ... Natarajan Kannan
    Research Article

    In eukaryotes, protein kinase signaling is regulated by a diverse array of post-translational modifications, including phosphorylation of Ser/Thr residues and oxidation of cysteine (Cys) residues. While regulation by activation segment phosphorylation of Ser/Thr residues is well understood, relatively little is known about how oxidation of cysteine residues modulate catalysis. In this study, we investigate redox regulation of the AMPK-related brain-selective kinases (BRSK) 1 and 2, and detail how broad catalytic activity is directly regulated through reversible oxidation and reduction of evolutionarily conserved Cys residues within the catalytic domain. We show that redox-dependent control of BRSKs is a dynamic and multilayered process involving oxidative modifications of several Cys residues, including the formation of intramolecular disulfide bonds involving a pair of Cys residues near the catalytic HRD motif and a highly conserved T-loop Cys with a BRSK-specific Cys within an unusual CPE motif at the end of the activation segment. Consistently, mutation of the CPE-Cys increases catalytic activity in vitro and drives phosphorylation of the BRSK substrate Tau in cells. Molecular modeling and molecular dynamics simulations indicate that oxidation of the CPE-Cys destabilizes a conserved salt bridge network critical for allosteric activation. The occurrence of spatially proximal Cys amino acids in diverse Ser/Thr protein kinase families suggests that disulfide-mediated control of catalytic activity may be a prevalent mechanism for regulation within the broader AMPK family.