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

Neural network explanation using inversion.

Emad W Saad1, Donald C Wunsch

  • 1Phantom Works, The Boeing Company, Seattle, WA 98124, United States. Emad.w.saad@boeing.com

Neural Networks : the Official Journal of the International Neural Network Society
|October 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

Current state of the art of new prostate MRI technologies and potential future developments.

BJR open·2026
Same author

Prostate Cancer: Current Status of Novel Molecular Imaging and Future Prospects.

European urology focus·2025
Same author

DeepART: Deep gradient-free local learning with adaptive resonance.

Neural networks : the official journal of the International Neural Network Society·2025
Same author

Adaptive Nussbaum Design for Nonholonomic Systems With Asymptotic Stabilization Against False Data Injection.

IEEE transactions on cybernetics·2025
Same author

The visualization of Orphadata neurology phenotypes.

Frontiers in digital health·2023
Same author

Subtypes of relapsing-remitting multiple sclerosis identified by network analysis.

Frontiers in digital health·2023
Same journal

Exploiting audio-visual modalities in videos: Object detection via multi-stage bilateral coupling network.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Reliability-aware modality completion with cross-modal distillation for federated learning with missing modalities.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

IGFD-Net: Illumination-guided frequency decoupling for polarization image fusion.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Multiple-Strategies dung beetle optimizer and its applications in engineering optimization and bankruptcy prediction.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Aggregating global-scale pixel-wise forgery cues within a graph.

Neural networks : the official journal of the International Neural Network Society·2026
Same journal

Finite-Time intermittent control for secure synchronization of Neutral-Type stochastic delayed neural networks under aperiodic DoS attacks.

Neural networks : the official journal of the International Neural Network Society·2026
See all related articles

This study introduces HYPINV, a novel algorithm for explaining artificial neural networks (ANNs) by inverting network outputs to generate hyperplane rules. HYPINV offers a fidelity-complexity tradeoff for rule extraction from ANNs.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Neural Networks

Background:

  • Artificial neural networks (ANNs) often lack interpretability, hindering their application in critical domains.
  • Existing methods for explaining ANN outputs have limitations in clarity and applicability.
  • Rule extraction from ANNs is crucial for understanding their decision-making processes.

Purpose of the Study:

  • To address the explanation capability drawback in artificial neural networks.
  • To introduce HYPINV, a new pedagogical algorithm for extracting rules from ANNs.
  • To present a novel network inversion technique for generating hyperplane rules.

Main Methods:

  • A survey of existing ANN explanation algorithms.
  • Development and presentation of the HYPINV algorithm based on network inversion.

Related Experiment Videos

  • Implementation of network inversion techniques including gradient descent and evolutionary algorithms.
  • Information theoretic analysis of rule extraction.

    • Main Results:

    • HYPINV successfully extracts rules in the form of hyperplanes from ANNs.
    • The algorithm demonstrates an ability to control the fidelity-complexity tradeoff.
    • HYPINV is shown to be effective on synthetic, aerospace, and benchmark problems.
    • It is the first pedagogical method to extract hyperplane rules from continuous or binary attribute neural networks.

    Conclusions:

    • HYPINV provides a valuable tool for enhancing the interpretability of artificial neural networks.
    • The fidelity-complexity tradeoff offers flexible rule generation for diverse applications.
    • Network inversion is a viable approach for pedagogical rule extraction in ANNs.