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

Calibratable Disambiguation Loss for Multi-Instance Partial-Label Learning.

Wei Tang, Yin-Fang Yang, Weijia Zhang

    IEEE Transactions on Pattern Analysis and Machine Intelligence
    |July 13, 2026
    PubMed
    Summary

    This study introduces a calibratable disambiguation loss (CDL) to improve multi-instance partial-label learning (MIPL) by enhancing classifier reliability. CDL significantly boosts classification accuracy and calibration in MIPL tasks.

    Related Concept Videos

    Multi-input and Multi-variable systems01:22

    Multi-input and Multi-variable systems

    Cruise control systems in cars are designed as multi-input systems to maintain a driver's desired speed while compensating for external disturbances such as changes in terrain. The block diagram for a cruise control system typically includes two main inputs: the desired speed set by the driver and any external disturbances, such as the incline of the road. By adjusting the engine throttle, the system maintains the vehicle's speed as close to the desired value as possible.
    In the absence of...
    Associative Learning01:27

    Associative Learning

    Associative learning is a fundamental concept in behavioral psychology, wherein a connection is established between two stimuli or events, leading to a learned response. This process is critical in understanding how behaviors are acquired and modified. Conditioning, the mechanism through which associations are formed, can be divided into two main types: classical conditioning and operant conditioning, each elucidating different aspects of associative learning.
    Classical conditioning, also known...

    You might also read

    Related Articles

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

    Sort by
    Same author

    Decouple then Converge: Handling Unknown Unlabeled Distributions in Long-Tailed Semi-Supervised Learning.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same author

    Privacy Preserving Decentralized Learning With Positive-Incentive Noise.

    IEEE transactions on pattern analysis and machine intelligence·2026
    Same author

    Decentralized Primal-Dual Optimization Without Global Lipschitz Continuity.

    IEEE transactions on neural networks and learning systems·2026
    Same author

    Evolutionary Dimension-Specific Feature Selection for Multi-Dimensional Classification.

    IEEE transactions on pattern analysis and machine intelligence·2025
    Same author

    Data Augmentation With Regularization for Multi-Labeled Complementary Label Learning.

    IEEE transactions on pattern analysis and machine intelligence·2025
    Same author

    Toward Few-Shot Learning in the Open World: A Review and Beyond.

    IEEE transactions on pattern analysis and machine intelligence·2025

    Area of Science:

    • Machine Learning
    • Artificial Intelligence
    • Computer Science

    Background:

    • Multi-instance partial-label learning (MIPL) combines multi-instance learning (MIL) and partial-label learning (PLL) for weakly supervised scenarios with inexact supervision.
    • Existing MIPL methods often exhibit poor calibration, leading to unreliable classifier performance.
    • Addressing poor calibration is crucial for enhancing the practical utility of MIPL frameworks.

    Purpose of the Study:

    • To propose a novel plug-and-play calibratable disambiguation loss (CDL) for MIPL.
    • To improve both classification accuracy and classifier calibration in MIPL.
    • To provide theoretical insights into the margin modulation and its effect on weight updates.

    Main Methods:

    • Developed a calibratable disambiguation loss (CDL) that modulates a disambiguation objective using a top-vs-competitor prediction margin.

    Related Experiment Videos

  • Introduced two variants of CDL: one focusing on candidate-level separation and another on candidate-vs-non-candidate suppression.
  • Analyzed CDL theoretically as a margin-modulated momentum-based disambiguation loss (MDL) and derived calibration bounds.

    • Main Results:

    • Experimental results demonstrate that CDL significantly enhances classification accuracy on benchmark and real-world MIPL datasets.
    • The proposed CDL method substantially reduces the expected calibration error, improving classifier reliability.
    • Evaluations on partial-label learning (PLL) adaptations also showed performance improvements.

    Conclusions:

    • The proposed calibratable disambiguation loss (CDL) effectively addresses the calibration issues in MIPL.
    • CDL offers a versatile and effective approach to improve both accuracy and reliability in weakly supervised learning.
    • The theoretical analysis provides a deeper understanding of margin-based disambiguation and calibration mechanisms.