Advertisement

BMC Neuroscience

, 16:P115 | Cite as

Influence of recurrent interactions on texture processing in networks with different visual map organizations

  • Hanna Kamyshanska
  • Dmitry Bibichkov
  • Matthias Kaschube
Poster presentation

Keywords

Orientation Selectivity Visual Texture Orientation Tuning Recurrent Connection Feed Forward Connection 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The functional architecture of the visual cortex displays marked differences across mammalian species: in stark contrast to primates, in which the preferred stimulus orientation forms an almost smooth map across the cortical surface, in rodents a 'salt-and-pepper' organization has been observed [1]. It is conceivable that the organization of preferred orientation has an impact on the processing of visual input. Recently [2], we found that in a biologically inspired object recognition system with a pure feed forward network architecture, smooth orientation maps outperform the salt-and-pepper organization in a texture recognition task. Here, we extend this work to study the effect of recurrent connections on neuronal response properties and texture recognition, comparing the two types of cortical architectures.

Our model is a recurrent single-layer rate network. The feed forward connections to each neuron are properly oriented Gabor-filters. Recurrent connections between two neurons depend both on the spatial distance between these neurons and on their orientation selectivities. Inspired by [3], the angle-dependent interaction function is weighted by the product of selectivities of both neurons and enables excitation between cells with similar preferred orientations and inhibition between cells with orthogonal ones. We also study the effects of purely distant-dependent recurrent connectivity of Mexican-hat type, with spatial extent related to local column spacing in case of smooth map layout. We design a network for the salt-and-pepper organization in an analogous way, assuming the same spatial extent of connectivity and its tuning-dependence as observed in mouse visual cortex [4].

Varying the strength and selectivity of recurrent versus feed-forward connections, we first explore the influence of recurrence on the orientation selectivity. For that purpose, we drive the network with oriented gratings to reconstruct the selectivity from the activities. In agreement with [5] we observe sharpening of the orientation tuning of the network by recurrent interactions, such that oriented stimuli can be well discriminated even with weak feed-forward tuning. We further study the role of recurrent connections in processing more complex stimuli. We present visual textures to the model, then feed the responses into a classifier (linear SVM) that learns to predict a class label. This allows us to study how differences in feed-forward and recurrent connections impact texture classification, and to compare the orientation-preference map and salt-and-pepper organization in texture recognition tasks.

References

  1. 1.
    Kaschube M: Neural maps versus salt-and-pepper organization in visual cortex. Curr Opin Neurobiol. 2014, 24: 95-102.PubMedCrossRefGoogle Scholar
  2. 2.
    Bauer F, Kaschube M: Processing textures in a smooth visual map and a salt-and-pepper organization. Bernstein Conference. 2013Google Scholar
  3. 3.
    Blumenfeld B, Bibitchkov D, & Tsodyks M: Neural network model of the primary visual cortex: From functional architecture to lateral connectivity and back. J Comput Neurosci. 2006, 20 (2): , 219-241.PubMedPubMedCentralCrossRefGoogle Scholar
  4. 4.
    Ko H, Hofer SB, Pichler B, Buchanan KA, Sjöström PJ, Mrsic-Flogel TD: Functional specificity of local synaptic connections in neocortical networks. Nature. 2011, 473 (7345): , 87-91.PubMedPubMedCentralCrossRefGoogle Scholar
  5. 5.
    Ben-Yishai R, Bar-Or RL, Sompolinsky H: Theory of orientation tuning in visual cortex. Proc Natl Acad Sci U S A. 1995, 92: , 3844-3848.PubMedPubMedCentralCrossRefGoogle Scholar

Copyright information

© Kamyshanska et al. 2015

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors and Affiliations

  • Hanna Kamyshanska
    • 1
  • Dmitry Bibichkov
    • 1
  • Matthias Kaschube
    • 1
  1. 1.Frankfurt Institute for Advanced Studies and Faculty of Computer Science and MathematicsJohann Wolfgang Goethe UniversityFrankfurt am MainGermany

Personalised recommendations