Jove
Visualize
Contact Us

Related Experiment Videos

Polynomial harmonic GMDH learning networks for time series modeling.

Nikolay Y Nikolaev1, Hitoshi Iba

  • 1Department of Computing, Goldsmiths College, University of London, New Cross, London SE14 6NW, UK. nikolaev@mcs.gold.ac.uk

Neural Networks : the Official Journal of the International Neural Network Society
|November 19, 2003
PubMed
Summary

This study introduces a novel polynomial harmonic neural network (phGMDH) that enhances time series modeling. The phGMDH model, utilizing polynomial activation and backpropagation, significantly outperforms 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

Pulse Detecting Genetic Circuit - A New Design Approach.

PloS one·2016
Same author

Evolving robust gene regulatory networks.

PloS one·2015
Same author

Reverse engineering gene regulatory network from microarray data using linear time-variant model.

BMC bioinformatics·2010
Same author

Prediction of cancer class with majority voting genetic programming classifier using gene expression data.

IEEE/ACM transactions on computational biology and bioinformatics·2009
Same author

Sequential Bayesian kernel modelling with non-Gaussian noise.

Neural networks : the official journal of the International Neural Network Society·2007
Same author

Inferring gene regulatory networks using differential evolution with local search heuristics.

IEEE/ACM transactions on computational biology and bioinformatics·2007
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

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Data Science

Background:

  • Traditional neural networks often struggle with complex, non-linear data patterns.
  • Existing Group Method of Data Handling (GMDH) algorithms have limitations in expressive power and performance.

Purpose of the Study:

  • To present a constructive approach for neural network modeling of polynomial harmonic functions.
  • To enhance neural network learning through novel compositional schemes and improved training algorithms.

Main Methods:

  • Developing a polynomial harmonic version of the GMDH algorithm (phGMDH).
  • Incorporating an analytical method for combining polynomial terms and harmonics.
  • Utilizing backpropagation for gradient descent weight optimization, initialized by least squares fitting.

Related Experiment Videos

Main Results:

  • The phGMDH model demonstrated superior performance on time series modeling tasks compared to standard GMDH, Neurofuzzy GMDH, and Multilayer Perceptron (MLP) networks.
  • The combination of polynomial terms and analytical harmonics extended the network's expressive power.
  • Backpropagation training further improved the performance of the higher-order polynomial networks.

Conclusions:

  • The proposed polynomial harmonic neural network approach offers significant advantages for time series modeling.
  • The integration of analytical harmonic components and advanced training methods enhances predictive accuracy.
  • phGMDH represents a powerful advancement in constructive neural network modeling.