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

A quantified sensitivity measure for multilayer perceptron to input perturbation.

Xiaoqin Zeng1, Daniel S Yeung

  • 1Department of Computing, Hong Kong Polytechnic University, Hong Kong. csxzeng@comp.polyu.edu.hk

Neural Computation
|February 20, 2003
PubMed
Summary

This study introduces a method to measure how sensitive a multilayer perceptron (MLP) neural network is to input changes. The proposed sensitivity measure helps evaluate MLP performance and understand its stability.

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

Site-Specific Hydroxide Formation and Corrosion on Mg Nanocrystals.

Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
Same author

Case Report: Ultrasound guided puncture for type 2 diabetes mellitus combined with psoas abscess-a report of two cases.

Frontiers in medicine·2026
Same author

Heterogeneous trajectory of cancer-related fatigue and its associated factors in patients with glioma during the perioperative period: a prospective longitudinal study.

Supportive care in cancer : official journal of the Multinational Association of Supportive Care in Cancer·2026
Same author

Chelating Anion-Mediated Solvation Structures for Rechargeable Magnesium Batteries.

Angewandte Chemie (International ed. in English)·2026
Same author

The structural characteristics, physicochemical properties, rheological behavior and anti-obesity effects of polysaccharides from Pueraria lobata: In-depth study focusing on the gut microbiota-hepatic FGF21 signaling.

Food research international (Ottawa, Ont.)·2026
Same author

Active learning design of bcc solid solution alloys with gigapascal strength and elemental metal-level ductility.

Proceedings of the National Academy of Sciences of the United States of America·2026

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Computational Neuroscience

Background:

  • Neural network sensitivity to input perturbations is critical for both theoretical understanding and practical applications.
  • Multilayer perceptrons (MLPs) are widely used feedforward neural networks, making their sensitivity analysis a significant research area.

Purpose of the Study:

  • To propose and define a novel approach for quantifying the sensitivity of multilayer perceptrons (MLPs) to input perturbations.
  • To develop algorithms for calculating MLP sensitivity based on its structural characteristics.

Main Methods:

  • A bottom-up approach was employed, starting with the sensitivity analysis of a single neuron.
  • Analytical expressions were derived for computing neuron sensitivity as a function of expected input deviation.

Related Experiment Videos

  • An algorithm was developed to compute the sensitivity of the entire MLP network.

    • Main Results:

    • Theoretical formulas for MLP sensitivity were derived and validated through computer simulations.
    • Experimental results showed good agreement with the derived theoretical formulas.
    • The proposed sensitivity measure effectively quantifies MLP output deviation concerning input patterns.

    Conclusions:

    • The developed sensitivity measure provides a valuable tool for evaluating MLP performance and robustness.
    • The approach offers a quantitative understanding of how input variations affect MLP outputs.
    • This work contributes to the theoretical foundation of neural network analysis and design.