Emergence of brain-like mirror-symmetric viewpoint tuning in convolutional neural networks

  1. Amirhossein Farzmahdi
  2. Wilbert Zarco
  3. Winrich A Freiwald
  4. Nikolaus Kriegeskorte
  5. Tal Golan  Is a corresponding author
  1. Rockefeller University, United States
  2. Columbia University, United States

Abstract

Primates can recognize objects despite 3D geometric variations such as in-depth rotations. The computational mechanisms that give rise to such invariances are yet to be fully understood. A curious case of partial invariance occurs in the macaque face-patch AL and in fully connected layers of deep convolutional networks in which neurons respond similarly to mirror-symmetric view (e.g., left and right profiles). Why does this tuning develop? Here, we propose a simple learning-driven explanation for mirror-symmetric viewpoint tuning. We show that mirror-symmetric viewpoint tuning for faces emerges in the fully connected layers of convolutional deep neural networks trained on object recognition tasks, even when the training dataset does not include faces. First, using 3D objects rendered from multiple views as test stimuli, we demonstrate that mirror-symmetric viewpoint tuning in convolutional neural network models is not unique to faces: it emerges for multiple object categories with bilateral symmetry. Second, we show why this invariance emerges in the models. Learning to discriminate among bilaterally symmetric object categories induces reflection-equivariant intermediate representations. AL-like mirror-symmetric tuning is achieved when such equivariant responses are spatially pooled by downstream units with sufficiently large receptive fields. These results explain how mirror-symmetric viewpoint tuning can emerge in neural networks, providing a theory of how they might emerge in the primate brain. Our theory predicts that mirror-symmetric viewpoint tuning can emerge as a consequence of exposure to bilaterally symmetric objects beyond the category of faces, and that it can generalize beyond previously experienced object categories.

Data availability

The stimulus set and the source code required for reproducing our results are available at https://gitfront.io/r/afarzmahdi/p666tmWy7YuY/AL-symmetry-manuscript-codes/.

The following data sets were generated

Article and author information

Author details

  1. Amirhossein Farzmahdi

    Laboratory of Neural Systems, Rockefeller University, New York, United States
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6926-546X
  2. Wilbert Zarco

    Laboratory of Neural Systems, Rockefeller University, New York, United States
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3599-0476
  3. Winrich A Freiwald

    Laboratory of Neural Systems, Rockefeller University, New York, United States
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8456-5030
  4. Nikolaus Kriegeskorte

    Zuckerman Mind Brain Behavior Institute, Columbia University, New York, United States
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7433-9005
  5. Tal Golan

    Zuckerman Mind Brain Behavior Institute, Columbia University, New York, United States
    For correspondence
    golan.neuro@bgu.ac.il
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7940-7473

Funding

National Eye Institute (R01EY021594)

  • Winrich A Freiwald

National Eye Institute (R01EY029998)

  • Winrich A Freiwald

National Institute of Neurological Disorders and Stroke (RF1NS128897)

  • Nikolaus Kriegeskorte

Naval Research Laboratory (N00014-20-1-2292)

  • Winrich A Freiwald

Charles H. Revson Foundation

  • Tal Golan

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

Copyright

© 2024, Farzmahdi 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

  • 878
    views
  • 149
    downloads
  • 0
    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. Amirhossein Farzmahdi
  2. Wilbert Zarco
  3. Winrich A Freiwald
  4. Nikolaus Kriegeskorte
  5. Tal Golan
(2024)
Emergence of brain-like mirror-symmetric viewpoint tuning in convolutional neural networks
eLife 13:e90256.
https://doi.org/10.7554/eLife.90256

Share this article

https://doi.org/10.7554/eLife.90256

Further reading

    1. Neuroscience
    Mihály Vöröslakos, Yunchang Zhang ... György Buzsáki
    Tools and Resources

    Brain states fluctuate between exploratory and consummatory phases of behavior. These state changes affect both internal computation and the organism’s responses to sensory inputs. Understanding neuronal mechanisms supporting exploratory and consummatory states and their switching requires experimental control of behavioral shifts and collecting sufficient amounts of brain data. To achieve this goal, we developed the ThermoMaze, which exploits the animal’s natural warmth-seeking homeostatic behavior. By decreasing the floor temperature and selectively heating unmarked areas, we observed that mice avoided the aversive state by exploring the maze and finding the warm spot. In its design, the ThermoMaze is analogous to the widely used water maze but without the inconvenience of a wet environment and, therefore, allows the collection of physiological data in many trials. We combined the ThermoMaze with electrophysiology recording, and report that spiking activity of hippocampal CA1 neurons during sharp-wave ripple events encode the position of mice. Thus, place-specific firing is not confined to locomotion and associated theta oscillations but persist during waking immobility and sleep at the same location. The ThermoMaze will allow for detailed studies of brain correlates of immobility, preparatory–consummatory transitions, and open new options for studying behavior-mediated temperature homeostasis.

    1. Neuroscience
    Sergio Casas-Tinto, Nuria Garcia-Guillen, María Losada-Perez
    Short Report

    As the global population ages, the prevalence of neurodegenerative disorders is fast increasing. This neurodegeneration as well as other central nervous system (CNS) injuries cause permanent disabilities. Thus, generation of new neurons is the rosetta stone in contemporary neuroscience. Glial cells support CNS homeostasis through evolutionary conserved mechanisms. Upon damage, glial cells activate an immune and inflammatory response to clear the injury site from debris and proliferate to restore cell number. This glial regenerative response (GRR) is mediated by the neuropil-associated glia (NG) in Drosophila, equivalent to vertebrate astrocytes, oligodendrocytes (OL), and oligodendrocyte progenitor cells (OPCs). Here, we examine the contribution of NG lineages and the GRR in response to injury. The results indicate that NG exchanges identities between ensheathing glia (EG) and astrocyte-like glia (ALG). Additionally, we found that NG cells undergo transdifferentiation to yield neurons. Moreover, this transdifferentiation increases in injury conditions. Thus, these data demonstrate that glial cells are able to generate new neurons through direct transdifferentiation. The present work makes a fundamental contribution to the CNS regeneration field and describes a new physiological mechanism to generate new neurons.