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

Hierarchical genetic algorithm for near optimal feedforward neural network design.

Gary Yen1, Haiming Lu

  • 1Intelligent Systems and Control Laboratory, School of Electrical and Computer Engineering, Stillwater, OK 74078, USA.

International Journal of Neural Systems
|February 20, 2002
PubMed
Summary

This study introduces a hierarchical genetic algorithm for designing multi-layer feedforward neural networks, optimizing both structure and weights. The novel approach demonstrates competitive or superior performance in chaotic time series prediction compared to existing methods.

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

Revisiting, Understanding, and Tailoring the Evolution in the Nature of Phase Transitions in Rare-Earth RE<sub>2</sub>In Alloys.

The journal of physical chemistry letters·2026
Same author

Source Apportionment of Nitrate Pollution in the Ling River Basin Based on Nitrogen and Oxygen Isotopes and Hydrochemical Analysis.

Bulletin of environmental contamination and toxicology·2026
Same author

Chitosan-Hyaluronic Acid Composite Hydrogel with Slow-Release Hydrogen Sulfide and Cerium Oxide for Multifunctional Synergy to Promote Healing of Infected Wounds.

International journal of nanomedicine·2026
Same author

[A coronary artery plaque segmentation method based on focal weighted accuracy loss function].

Sheng wu yi xue gong cheng xue za zhi = Journal of biomedical engineering = Shengwu yixue gongchengxue zazhi·2025
Same author

Impact of Tobacco Exposure on Low Back Pain-Related Disability-Adjusted Life Years in China: Trends From 1990-2021 and Projections for the Next 15 years Using ARIMA Modeling.

Journal of pain research·2025
Same author

Tuning of the Electronic and Magnetic Properties of GaN Monolayers via Doping with Lanthanide Atoms and by Applying Biaxial Strain.

Nanomaterials (Basel, Switzerland)·2025

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computational Neuroscience

Background:

  • Traditional genetic algorithms for neural network design have limitations.
  • Existing methods often struggle with network topology and parameter optimization simultaneously.
  • A feasibility check is a known deficiency in conventional genetic algorithm-based neural network designs.

Purpose of the Study:

  • To propose a novel hierarchical genetic algorithm for designing multi-layer feedforward neural networks.
  • To optimize both the topology and weighting parameters of neural networks concurrently.
  • To address deficiencies in traditional genetic algorithm-based neural network design procedures.

Main Methods:

  • Utilizing a hierarchical genetic algorithm to evolve neural network topology and weights.

Related Experiment Videos

  • Implementing a multi-objective cost function for simultaneous optimization of performance and topology.
  • Applying a linear weight combination for decision-making to derive an approximated Pareto optimal solution set.

    • Main Results:

    • The proposed approach designed neural networks that are competitive or superior to traditional algorithms.
    • Performance was evaluated using the prediction of Mackey Glass chaotic time series.
    • The method successfully optimized neural network performance and topology simultaneously.

    Conclusions:

    • The hierarchical genetic algorithm offers an effective method for designing advanced neural networks.
    • The approach provides a robust solution for the two-objective optimization problem in neural network design.
    • This method shows significant promise for applications in complex time series prediction.