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

Dimensional reduction for reward-based learning.

Christian D Swinehart1, L F Abbott

  • 1Department of Biology, Volen Center for Complex Systems, Brandeis University, Waltham, MA 02454-9110, USA.

Network (Bristol, England)
|December 13, 2006
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

Neuronal calcium spikes enable vector inversion in the Drosophila brain.

Cell·2025
Same author

Transfer of graded information through gated receptivity to widely broadcast signals.

bioRxiv : the preprint server for biology·2025
Same author

Motor cortex flexibly deploys a high-dimensional repertoire of subskills.

bioRxiv : the preprint server for biology·2025
Same author

Associative synaptic plasticity creates dynamic persistent activity.

bioRxiv : the preprint server for biology·2025
Same author

Connectome analysis of a cerebellum-like circuit for sensory prediction.

bioRxiv : the preprint server for biology·2025
Same author

An end-to-end model of active electrosensation.

Current biology : CB·2025
Same journal

Enhancing IoT security: A Creative Swagger Optimization algorithm for DDoS defence.

Network (Bristol, England)·2026
Same journal

Parametric optimization for electrical discharge diamond grinding (EDDG) system using dual approach.

Network (Bristol, England)·2025
Same journal

A novel lung cancer diagnosis model using hybrid convolution (2D/3D)-based adaptive DenseUnet with attention mechanism.

Network (Bristol, England)·2025
Same journal

Hybrid optimization enabled Eff-FDMNet for Parkinson's disease detection and classification in federated learning.

Network (Bristol, England)·2025
Same journal

AI-driven plant disease detection with tailored convolutional neural network.

Network (Bristol, England)·2025
Same journal

Layer modified residual Unet++ for speech enhancement using Aquila Black widow optimizer algorithm.

Network (Bristol, England)·2025
See all related articles

Hebbian plasticity in supervisor circuits enables efficient reinforcement learning in large neural networks by reducing search space. This method facilitates rapid function approximation and allows networks to eventually learn independently.

Area of Science:

  • Computational Neuroscience
  • Machine Learning
  • Artificial Intelligence

Background:

  • Optimizing large neural networks using reward-based learning is computationally challenging due to vast parameter spaces.
  • Existing reinforcement learning methods struggle with efficiency in large-scale network optimization.
  • The need for effective parameter space exploration in complex artificial neural systems is critical.

Purpose of the Study:

  • To investigate a novel approach for efficient reinforcement learning in large neural networks.
  • To demonstrate how Hebbian plasticity in supervisor circuits can facilitate large-scale network optimization.
  • To explore the role of reciprocal connections in supervised learning dynamics.

Main Methods:

  • Utilized Hebbian forms of synaptic plasticity to establish connections between a supervisor circuit and a controlled network.

Related Experiment Videos

  • Implemented reciprocal connections between supervisor units and the network being controlled.
  • Employed a reinforcement-based learning procedure following the setup of Hebbian connections for a function approximation task.

    • Main Results:

    • Hebbian plasticity effectively reduced the dimensionality of the parameter search space for the controlled network.
    • Reciprocal connections between supervisor and network were crucial for efficient learning.
    • Rapid learning was achieved in a function approximation task, demonstrating the efficacy of the proposed method.

    Conclusions:

    • Hebbian plasticity in supervisor circuits offers an efficient mechanism for reinforcement learning in large neural networks.
    • The proposed method significantly enhances the speed and efficiency of learning complex tasks.
    • Hebbian plasticity within the supervised network ultimately enables autonomous task performance without supervisor input.