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

Simple models for reading neuronal population codes

H S Seung1, H Sompolinsky

  • 1AT&T Bell Laboratories, Murray Hill, NJ 07974.

Proceedings of the National Academy of Sciences of the United States of America
|November 15, 1993
PubMed
Summary
This summary is machine-generated.

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

Integrating EM and Patch-seq data: Synaptic connectivity and target specificity of predicted Sst transcriptomic types.

bioRxiv : the preprint server for biology·2023
Same author

Young adult-born neurons improve odor coding by mitral cells.

Nature communications·2020
Same author

Dynamics of random neural networks with bistable units.

Physical review. E, Statistical, nonlinear, and soft matter physics·2015
Same author

Spontaneous eye movements in goldfish: oculomotor integrator performance, plasticity, and dependence on visual feedback.

Vision research·2004
Same author

Robust persistent neural activity in a model integrator with multiple hysteretic dendrites per neuron.

Cerebral cortex (New York, N.Y. : 1991)·2003
Same author

Linear regression of eye velocity on eye position and head velocity suggests a common oculomotor neural integrator.

Journal of neurophysiology·2002

This study models neural networks for extracting sensory information from neuronal populations. Broad tuning curves improve direction estimation, and perceptual learning transfer shows nonmonotonicity in neural network models.

Area of Science:

  • Computational Neuroscience
  • Neural Network Modeling
  • Sensory Information Processing

Background:

  • Sensory information is often distributed across neuronal populations in neural systems.
  • Understanding how neural networks extract this distributed information is crucial for neuroscience.
  • Previous models often simplify the complexity of neuronal responses and network architectures.

Purpose of the Study:

  • To develop and evaluate simple neural network models for extracting directional sensory information.
  • To compare the performance of these models against optimal maximum likelihood estimation in psychophysical tasks.
  • To investigate direction estimation and discrimination using population vector and adaptive network models.

Main Methods:

  • Simulated stochastic responses of sensory neuron populations tuned to directional stimuli.

Related Experiment Videos

  • Developed a linear network model computing a population vector for direction estimation.
  • Analyzed perceptron and adaptive two-layer networks for direction discrimination, calculating error rates and training transfer.

    • Main Results:

    • Population vector performance for direction estimation is optimal with broad tuning curves, influenced by background activity.
    • Performance gap between population vector and maximum likelihood estimation narrows with broader tuning.
    • Perceptual learning transfer in threshold linear networks is nonmonotonic, with performance peaking at intermediate angles.

    Conclusions:

    • Neural network models can effectively extract distributed sensory information, with performance contingent on tuning width and network architecture.
    • Broad tuning curves enhance direction estimation accuracy in population vector models.
    • The nonmonotonic transfer of perceptual learning in adaptive networks offers a testable psychophysical prediction for neural models.