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

Adaptive probabilistic neural networks for pattern classification in time-varying environment.

Leszek Rutkowski1

  • 1Department of Computer Engineering, Technical University in Czestochowa. lrutko@kik.pcz.czest.pl

IEEE Transactions on Neural Networks
|October 6, 2004
PubMed
Summary

This study introduces novel probabilistic neural networks (PNNs) for pattern classification in non-stationary environments. These PNNs adapt to changing data distributions, offering improved prediction and classification accuracy over time.

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

ARKG: Adversarially Residual Knowledge Generalization to Open-Set Domain Adaptation.

IEEE transactions on neural networks and learning systems·2026
Same author

Random Spatiotemporal Sampled-Data Control for Reaction-Diffusion Neural Networks With Dwell-Time-Based Sojourn Probability.

IEEE transactions on cybernetics·2026
Same author

Distributed Topology Reconfiguration for Open Multiagent Systems via Algebraic Connectivity and Kirchhoff Index.

IEEE transactions on cybernetics·2026
Same author

Stochastic-Sampling-Based Event-Triggered Control for Switching Reaction-Diffusion Neural Networks.

IEEE transactions on cybernetics·2026
Same author

DM-CFO: A Diffusion Model for Compositional 3D Tooth Generation With Collision-Free Optimization.

IEEE transactions on visualization and computer graphics·2026
Same author

Observer-Based Adaptive Neural Sliding Mode Control of Fuzzy Systems With Sojourn-Probability-Based Multimode Attacks.

IEEE transactions on cybernetics·2025

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Pattern Recognition

Background:

  • Traditional pattern classification methods often struggle in non-stationary environments where data distributions change over time.
  • Adapting to time-varying probability distributions is crucial for robust pattern recognition systems.

Purpose of the Study:

  • To propose a new class of probabilistic neural networks (PNNs) designed to operate effectively in non-stationary environments.
  • To formulate pattern classification in non-stationary settings as a prediction problem.
  • To establish theoretical foundations for PNN optimality in time-varying scenarios.

Main Methods:

  • Formulating pattern classification in non-stationary environments as a prediction problem.
  • Designing probabilistic neural networks capable of handling time-varying probability distributions.

Related Experiment Videos

  • Developing definitions of optimality for PNNs in time-varying environments.
  • Proving asymptotic convergence to Bayes-optimal decision surfaces.
  • Investigating the convergence speed of the proposed PNNs.
  • Implementing PNNs using Parzen kernels and multivariate Hermite series.

    • Main Results:

    • A novel class of probabilistic neural networks (PNNs) for non-stationary environments has been developed.
    • The proposed PNNs are shown to asymptotically approach the Bayes-optimal (time-varying) decision surface.
    • The convergence properties of these PNNs have been investigated.
    • Detailed designs for PNNs utilizing Parzen kernels and multivariate Hermite series are presented.

    Conclusions:

    • The developed PNNs offer a robust solution for pattern classification in dynamic, non-stationary environments.
    • The theoretical framework and practical designs presented advance the field of adaptive pattern recognition.
    • These PNNs demonstrate significant potential for applications requiring real-time adaptation to changing data characteristics.