Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Learning distinct and complementary feature selectivities from natural colour videos.

Wolfgang Einhäuser1, Christoph Kayser, Konrad P Körding

  • 1Institute of Neuroinformatics (UNI/ETH Zürich), Zürich, Switzerland. weinhaeu@ini.phys.ethz.ch

Reviews in the Neurosciences
|August 22, 2003
PubMed
Summary

This study shows how neural networks can develop specialized neurons for processing visual information. By optimizing for stable and decorrelated responses, neurons naturally segregate into color-selective and orientation-selective groups.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Spontaneous perceptual reversals are accompanied by systematic changes in pupil size but not respiration phase.

Scientific reports·2026
Same author

Physical contact reveals a hidden layer of cortical architecture.

bioRxiv : the preprint server for biology·2026
Same author

Modeling attention and binding in the brain through bidirectional recurrent gating.

Nature communications·2026
Same author

Mid-superior temporal sulcus encodes spatial context and behavioral state in freely moving macaques.

bioRxiv : the preprint server for biology·2026
Same author

Sound lateralization ability is affected by saccade direction but not eye movement-related eardrum oscillations.

Journal of neurophysiology·2026
Same author

The alignment of respiration to sensory-motor events is shaped by expected effort.

iScience·2026

Area of Science:

  • Computational neuroscience
  • Artificial intelligence
  • Visual processing

Background:

  • Biological and artificial neural networks often require parallel feature extraction using distinct neuronal populations.
  • Homogeneous neuronal populations must segregate into specialized groups to achieve distinct response properties.

Purpose of the Study:

  • To investigate how initially homogeneous neuronal populations segregate into distinct functional groups.
  • To explore the emergence of complementary response properties in neural networks through self-organization.

Main Methods:

  • Training a neural network with objectives for optimally stable and decorrelated neuronal responses.
  • Utilizing a color image sequence from a freely behaving cat's head-mounted camera.
  • Analyzing the response properties of emergent neuronal subpopulations.

Related Experiment Videos

Main Results:

  • The objective of stable responses led to the development of color-selective neurons.
  • Adding a decorrelation objective resulted in a subpopulation of achromatic neurons.
  • Color-selective neurons exhibited non-orientation tuning, while achromatic neurons were orientation-tuned.

Conclusions:

  • The proposed training objectives successfully induced segregation of neurons into complementary populations.
  • This self-organization mechanism explains the parallel processing of distinct stimulus features like color and orientation.