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

Performance Bounds for Single Layer Threshold Networks when Tracking a Drifting Adversary.

Anthony Kuh1, Xiaodong Tian

  • 1University of Hawaii at Manoa, USA

Neural Networks : the Official Journal of the International Neural Network Society
|July 1, 1997
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

Optimizing Postoperative Prognostication in Duodenal Adenocarcinoma: A Multicenter Study with External Validation.

The oncologist·2026
Same author

SGLT2 inhibition induces autophagic flux blockade and sensitizes pancreatic cancer to EGFR-targeted therapy.

Cellular oncology (Dordrecht, Netherlands)·2026
Same author

Histone lactylation boosted SET8 potentiates carcinogenesis and angiogenesis of pancreatic ductal adenocarcinoma by cooperating with MTA1/NuRD complex.

Journal of translational medicine·2026
Same author

Tumor infiltrating B cells and tertiary lymphoid structures in pancreatic cancer prognosis and therapy.

iScience·2026
Same author

Exosomal miR-4644 Targets SPRY3 to Promote Proliferation and Invasion of Pancreatic Cancer.

Cancer medicine·2026
Same author

STC2 mediates hypoxia-induced immunosuppression and tumorigenesis in hepatocellular carcinoma.

Cellular signalling·2026

This study analyzes tracking algorithm errors against worst-case adversaries. It establishes new upper bounds for generalization error in perceptron and least mean square (LMS) trackers under slow weight changes.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Signal Processing

Background:

  • System identification models are crucial for understanding algorithm performance.
  • The drift problem, where target weights change slowly, poses challenges for tracking algorithms.
  • Previous research analyzed random unbiased drifting adversaries.

Purpose of the Study:

  • To determine upper bounds for generalization error in three tracking algorithms facing a worst-case adversary.
  • To analyze the impact of a worst-case drifting adversary on system identification.
  • To investigate the performance of optimal conservative, perceptron, and least mean square (LMS) trackers.

Main Methods:

  • Utilizing a system identification model with single-layer threshold networks for both target and tracking.

Related Experiment Videos

  • Analyzing the generalization error under a worst-case drifting adversary scenario.
  • Deriving analytical upper bounds for small drift rates (gamma).

    • Main Results:

    • Established upper bounds for optimal conservative tracker (2gamma/cos(gammapi)), perceptron tracker (2gammann/), and LMS tracker (gamma(2n + 2.5)).
    • Validated analytical findings through simulation results.
    • Demonstrated that derived bounds are tight for perceptron and LMS trackers at small drift rates.

    Conclusions:

    • The analysis provides crucial insights into the robustness of tracking algorithms against adversarial conditions.
    • The derived bounds offer a theoretical foundation for evaluating and improving tracking algorithm performance.
    • Further discussion addresses the influence of noise and input characteristics on algorithm behavior.