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

Supervised and unsupervised learning with two sites of synaptic integration.

K P Körding1, P König

  • 1Institute of Neuroinformatics, ETH/UNI Zürich, Winterthurerstr. 190, 8057 Zürich, Switzerland. koerding@ini.phys.ethz.ch

Journal of Computational Neuroscience
|January 18, 2002
PubMed
Summary

This study proposes a biologically plausible neural network learning model. By utilizing two neuronal variables, it implements abstract learning rules like backpropagation, enhancing biological realism.

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

Spirit of hope: Diminished hopelessness mediates the serial relation between spiritual experiences, reduced stress, and positive affect in transdiagnostic and healthy individuals.

Social science & medicine (1982)·2025
Same author

[New working time models-what is possible and how?]

Urologie (Heidelberg, Germany)·2025
Same author

Death in four RHDV2-vaccinated pet rabbits due to rabbit haemorrhagic disease virus 2 (RHDV2).

The Journal of small animal practice·2021
Same author

Toward Perceiving Robots as Humans: Three Handshake Models Face the Turing-Like Handshake Test.

IEEE transactions on haptics·2016
Same author

Effects of contextual information and stimulus ambiguity on overt visual sampling behavior.

Vision research·2015
Same author

Abstracts of Presentations at the International Conference on Basic and Clinical Multimodal Imaging (BaCI), a Joint Conference of the International Society for Neuroimaging in Psychiatry (ISNIP), the International Society for Functional Source Imaging (ISFSI), the International Society for Bioelectromagnetism (ISBEM), the International Society for Brain Electromagnetic Topography (ISBET), and the EEG and Clinical Neuroscience Society (ECNS), in Geneva, Switzerland, September 5-8, 2013.

Clinical EEG and neuroscience·2013

Area of Science:

  • Neuroscience
  • Computational Neuroscience
  • Artificial Intelligence

Background:

  • Neural network learning rules often rely on abstract objective functions and gradient ascent for weight optimization.
  • Standard models require neurons to store two variables: activity (sensory information) and a learning variable (objective function derivative).
  • The physiological basis for these dual variables in neurons is unclear, leading to questions about biological realism.

Purpose of the Study:

  • To investigate if recent findings on cortical pyramidal neuron properties can provide a physiological basis for abstract neural network learning rules.
  • To demonstrate the implementation of established learning algorithms using these physiologically inspired principles.

Main Methods:

  • Discussed the dual-site synaptic integration properties of cortical pyramidal neurons (basal and apical dendrites).

Related Experiment Videos

  • Implemented the backpropagation of error algorithm within a framework that utilizes these two neuronal variables.
  • Demonstrated a specific self-supervised learning algorithm using the proposed physiological basis.

    • Main Results:

    • Cortical pyramidal neurons, with distinct basal and apical dendritic integration sites, can naturally accommodate the two variables required by abstract learning rules.
    • The implementation of backpropagation and a self-supervised algorithm showed feasibility using this two-variable neuron model.
    • This approach allows for the incorporation of physiologically inspired properties into neural networks with minimal added complexity.

    Conclusions:

    • Recent discoveries about pyramidal neuron structure offer a potential physiological foundation for abstract neural network learning rules.
    • This biologically grounded approach enables more realistic neural network models without significant increases in computational complexity.
    • The findings suggest a path towards more biologically plausible artificial intelligence systems.