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 Collaborative Neurodynamic Approach to Sharpness-Aware Minimization for Improving Generalization of Deep Learning.

Dan Su, Jie Han, Chunhua Yang

    IEEE Transactions on Cybernetics
    |July 15, 2026
    PubMed
    Summary

    This study introduces CNSAM, a novel approach to improve deep neural network generalization by enhancing sharpness-aware minimization (SAM). CNSAM diversifies exploration for flatter loss landscapes, boosting model performance.

    Related Concept Videos

    Maximizing the Directional Derivative01:25

    Maximizing the Directional Derivative

    The directional derivative is a central concept in multivariable calculus that describes how a function changes at a given point when moving in a specified direction. This direction is represented by a unit vector, ensuring that only the orientation influences the rate of change. By varying the direction, different rates of change can be observed, demonstrating that the directional derivative depends strongly on the chosen direction.The directional derivative is computed using the gradient...

    You might also read

    Related Articles

    Articles linked to this work by shared authors, journal, and citation graph.

    Sort by
    Same author

    Reprogramming the spleen-brain axis: Splenic monocyte/macrophage-mediated nanotherapy for ischemic stroke.

    Acta pharmaceutica Sinica. B·2026
    Same author

    Enhancing interpretable soft sensing with embedded hybrid modeling: the GraphTrans approach for industrial processes.

    ISA transactions·2026
    Same author

    Endometrial Cancer of the Uterine Isthmus: A Cohort Study on Clinicopathological Features and Prognosis.

    Cancer medicine·2026
    Same author

    Dietary Fermented Chinese Chive Juice Improves Growth Performance and Reshapes the Fresh Meat Volatile Flavor Profile of Small-Tailed Han Sheep.

    Animals : an open access journal from MDPI·2026
    Same author

    Distribution of divalent small interfering RNA into neurons of sensory ganglia produces selective, durable knockdown of Nav1.7 and strong analgesia.

    Pain·2026
    Same author

    Epidemiology of respiratory infectious diseases after the relaxation of COVID-19 policies: a retrospective study in Xiamen, China.

    Frontiers in public health·2026

    Area of Science:

    • Deep Learning
    • Machine Learning
    • Artificial Intelligence

    Background:

    • Generalization is a key challenge in deep neural networks (DNNs), particularly for overparameterized models.
    • Sharpness-aware minimization (SAM) improves generalization by seeking flat regions in the loss landscape.
    • The original SAM's reliance on local linearization limits its generalization performance and introduces sensitivity to anchor points.

    Purpose of the Study:

    • To propose a new collaborative neurodynamic approach to SAM (CNSAM) to overcome limitations of the original SAM algorithm.
    • To enhance the generalization capabilities of deep neural networks.

    Main Methods:

    • CNSAM employs multiple agents for diversified exploration of the loss landscape.
    • This approach aims to identify flatter regions more effectively than traditional SAM.

    Related Experiment Videos

  • The method was tested on image classification and natural language processing (NLP) tasks.

    • Main Results:

    • CNSAM demonstrated consistent improvements in generalization across experiments.
    • The proposed method achieved lower test loss and higher accuracy compared to SAM.
    • CNSAM outperformed other related learning algorithms in extensive evaluations.

    Conclusions:

    • CNSAM offers a more robust and effective method for improving DNN generalization.
    • The collaborative neurodynamic strategy enhances the benefits of sharpness-aware minimization.
    • This approach shows significant promise for advancing deep learning model performance.