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 constrained optimization approach to preserving prior knowledge during incremental training.

Silvia Ferrari1, Mark Jensenius

  • 1Department of Mechanical Engineering and Materials Science, Duke University, Durham, NC 27708-0005 USA. sferrari@duke.edu

IEEE Transactions on Neural Networks
|June 11, 2008
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

Extracorporeal Cytokine Hemadsorption for the Treatment of Severe Steroid Refractory Cytokine Release Syndrome Secondary to T-Cell Immunotherapy for B-Cell Malignancies: A Case Series.

Hematological oncology·2026
Same author

Social inclusion inequalities between individuals with mental disorders and the general population: a systematic review and meta-analysis.

Epidemiology and psychiatric sciences·2026
Same author

Clinical, economic, and organisational impact of the novel AAP 2022 neonatal hyperbilirubinemia guidelines in an Italian third-level neonatology unit.

European journal of pediatrics·2026
Same author

"Okay, So It's Not Me"-The <i>Extra-Fatigue</i> of Formal and Informal Reporting of Sexual Harassment in Academia.

International journal of environmental research and public health·2026
Same author

Case Report: Hermansky-Pudlak syndrome type 1 with pulmonary involvement in a 62-year-old Caucasian woman with rheumatoid arthritis.

Frontiers in immunology·2026
Same author

Cannabidiol for psychotic disorders: A meta-analysis of randomized controlled trials.

Journal of psychopharmacology (Oxford, England)·2026
Same journal

Universal perceptron and DNA-like learning algorithm for binary neural networks: LSBF and PBF implementations.

IEEE transactions on neural networks·2013
Same journal

Guest editorial: special section on white box nonlinear prediction models.

IEEE transactions on neural networks·2011
Same journal

Data-based fault-tolerant control of high-speed trains with traction/braking notch nonlinearities and actuator failures.

IEEE transactions on neural networks·2011
Same journal

Guest editorial: special section on data-based control, modeling, and optimization.

IEEE transactions on neural networks·2011
Same journal

Neural network-based multiple robot simultaneous localization and mapping.

IEEE transactions on neural networks·2011
Same journal

Data-driven model-free adaptive control for a class of MIMO nonlinear discrete-time systems.

IEEE transactions on neural networks·2011
See all related articles

This study introduces a novel neural network training method that preserves prior knowledge (long-term memory) during incremental learning (short-term memory) using constrained optimization. This approach ensures that previously learned information is retained while adapting to new data.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Control Systems Engineering

Background:

  • Neural networks require effective methods for incremental learning to adapt to new information over time.
  • Preserving existing knowledge during retraining is crucial for stability and performance in dynamic environments.
  • Supervised learning often struggles to balance learning new tasks with retaining previously acquired knowledge.

Purpose of the Study:

  • To develop a supervised neural network training technique that preserves prior knowledge during incremental training.
  • To formulate online learning as a constrained error minimization problem.
  • To demonstrate the efficacy of this method in adaptive critic designs, specifically for flight control.

Main Methods:

  • A supervised neural network training technique based on constrained optimization.

Related Experiment Videos

  • Expressing prior knowledge as equality constraints derived from an algebraic training technique (long-term memory).
  • Formulating incremental training as an error minimization problem subject to equality constraints, utilizing an adjoined error gradient and backpropagation.

    • Main Results:

    • The proposed constrained training method accurately preserves long-term memory (prior knowledge) during repeated incremental training sessions.
    • The technique effectively assimilates new short-term memories (new data) without compromising established performance baselines.
    • Analytical and numerical results confirm the successful preservation of prior knowledge in an adaptive critic flight controller.

    Conclusions:

    • The developed constrained optimization technique provides a robust solution for incremental neural network training while preserving essential prior knowledge.
    • This method is particularly applicable to adaptive critic designs and real-world control systems requiring continuous adaptation.
    • The approach offers a significant advancement in lifelong learning for artificial intelligence systems.