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

Evolutionary credit apportionment and its application to time-dependent neural processing

R Smalz1, M Conrad

  • 1Department of Computer Science, Wayne State University, Detroit, MI 48202, USA.

Bio Systems
|January 1, 1995
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

ERPs reveal an iconic relation between sublexical phonology and affective meaning.

Cognition·2022
Same author

Adverse events during nursing care procedure in intensive care unit: The PREVENIR study.

Intensive & critical care nursing·2020
Same author

Correction to: Functional outcomes in adult patients with herpes simplex encephalitis admitted to the ICU: a multicenter cohort study.

Intensive care medicine·2019
Same author

Revealing interfacial disorder at the growth-front of thick many-layer epitaxial graphene on SiC: a complementary neutron and X-ray scattering investigation.

Nanoscale·2019
Same author

Functional outcomes in adult patients with herpes simplex encephalitis admitted to the ICU: a multicenter cohort study.

Intensive care medicine·2019
Same author

Corrigendum to "European contribution to the study of ROS: A summary of the findings and prospects for the future from the COST action BM1203 (EU-ROS)" [Redox Biol. 13 (2017) 94-162].

Redox biology·2017
Same journal

Ruliological Resilience: Pattern Restoration and Robustness in Wolfram Patterns. A Basis for Regeneration, Not Just in Cone Shells?

Bio Systems·2026
Same journal

The quantum-to-classical transducer: A thermodynamic and quantum mechanical framework for the emergence of bioenergetics.

Bio Systems·2026
Same journal

Forward-backward gene expression binarization for boolean state inference over a known regulatory network.

Bio Systems·2026
Same journal

Partial-label metric ceilings for evaluating gene regulatory networks inferred from single-cell foundation models.

Bio Systems·2026
Same journal

The impedance mismatch theory: A non-equilibrium thermodynamic framework for a shared energetic stress pathway in neurodegeneration.

Bio Systems·2026
Same journal

Immune signal-status misclassification: A theoretical framework for biological status assignment and failed status resolution.

Bio Systems·2026
See all related articles

This study introduces a novel training method for recurrent neural networks using evolutionary principles and credit assignment. It enhances temporal processing by enabling neurons to share information and adapt parameters based on collective performance.

Area of Science:

  • Neuroscience
  • Artificial Intelligence
  • Computational Neuroscience

Background:

  • Recurrent neural networks (RNNs) are crucial for temporal processing.
  • Existing training methods face challenges in optimizing complex network dynamics.
  • Evolutionary computation offers potential for novel network training paradigms.

Purpose of the Study:

  • To introduce and evaluate a new training approach for recurrent neural networks.
  • To address temporal neural processing challenges using a hybrid evolutionary and credit assignment method.
  • To compare the proposed algorithm with traditional genetic algorithms.

Main Methods:

  • A novel training algorithm combining Darwinian variation and selection with credit apportionment.
  • Utilizing competing networks with interconnections for neuron output sharing.

Related Experiment Videos

  • Assigning credit to individual neurons based on group performance and parameter inheritance.
  • Demonstrating the algorithm with connectionist-type units on temporal tasks.

    • Main Results:

    • The proposed method effectively trains recurrent neural networks for temporal processing.
    • Neuron performance and parameter sharing across networks demonstrated improved adaptation.
    • Comparison with genetic algorithms indicated competitive or superior performance.
    • Analysis of neuron firing behavior revealed insights into network optimization.

    Conclusions:

    • The developed training approach offers a promising alternative for RNNs in temporal tasks.
    • The integration of evolutionary principles and credit assignment facilitates robust neural network learning.
    • This method provides a scalable and adaptable framework for complex computational neuroscience problems.