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 signal-flow-graph approach to on-line gradient calculation.

P Campolucci1, A Uncini, F Piazza

  • 1Dipartimento di Elettronica ed Automatica, Università di Ancona, 60121 Ancona, Italy. campoluc@tiscalinet.it

Neural Computation
|August 23, 2000
PubMed
Summary

This study introduces a novel gradient computation method using signal flow graphs (SFGs) for dynamic adaptive systems. This approach simplifies gradient calculation for efficient on-line and batch learning in complex systems.

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

Chimeric antigen receptor T cell therapy cardiotoxicity: a narrative review.

Cardio-oncology (London, England)·2026
Same author

Evidence for the Collective Nature of Radial Flow in Pb+Pb Collisions with the ATLAS Detector.

Physical review letters·2026
Same author

Evidence for the Dimuon Decay of the Higgs Boson in pp Collisions with the ATLAS Detector.

Physical review letters·2025
Same author

Evidence for Longitudinally Polarized W Bosons in the Electroweak Production of Same-Sign W Boson Pairs in Association with Two Jets in pp Collisions at sqrt[s]=13  TeV with the ATLAS Detector.

Physical review letters·2025
Same author

Isolated shoulder palsy due to cortical infarction: localisation and electrophysiological correlates of recovery.

BMJ case reports·2025
Same author

Observation of tt[over ¯] Production in Pb+Pb Collisions at sqrt[s_{NN}]=5.02  TeV with the ATLAS Detector.

Physical review letters·2025

Area of Science:

  • Dynamic Systems and Control
  • Computational Neuroscience
  • Machine Learning

Background:

  • Nonlinear dynamic adaptive systems, including recurrent neural networks, are often represented by signal flow graphs (SFGs).
  • SFGs offer a powerful, circuit-like analogy for complex systems but are underutilized for rigorous computation.
  • Existing methods for gradient computation can be complex and computationally intensive.

Purpose of the Study:

  • To derive a method for on-line and batch-backward gradient computation using SFG representation theory.
  • To enable efficient gradient calculation for system parameters in dynamic adaptive systems.
  • To provide a computationally tractable approach for sensitivity analysis and system learning.

Main Methods:

  • Utilized signal flow graph (SFG) representation theory and its properties.

Related Experiment Videos

  • Derived gradient computation by analyzing the original SFG and its adjoint.
  • Employed discrete-time notation, applicable to continuous-time systems as well.

    • Main Results:

    • Developed a straightforward method for gradient computation without complex chain rule expansions.
    • Demonstrated the applicability to any causal, nonlinear, time-variant dynamic system representable by an SFG.
    • Showcased the method's utility for both off-line and on-line learning scenarios.

    Conclusions:

    • The SFG-based gradient computation method offers a simplified and efficient approach for dynamic systems.
    • This method is particularly valuable for on-line learning in applications like signal processing, control, and equalization.
    • The adjoint SFG analysis provides a robust framework for sensitivity analysis and adaptive system training.