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

An incremental learning algorithm with confidence estimation for automated identification of NDE signals.

Robi Polikar1, Lalita Udpa, Satish Udpa

  • 1Department of Electrical and Computer Engineering, Rowan University, Glassboro, NJ 08028, USA. polikar@rowan.edu

IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control
|September 4, 2004
PubMed
Summary

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

Machine learning identifies clinical and sociodemographic factors contributing to post-COVID hospitalization and cardiovascular risk in patients with peripheral artery disease.

Scientific reports·2025
Same author

DGX: Uncovering General Behavior of Deep Graph Models With Model-Level Explanation.

IEEE transactions on computational biology and bioinformatics·2025
Same author

iSeqSearch: incremental protein search for iBlast/iMMSeqs2/iDiamond.

PeerJ·2025
Same author

The Naïve Bayes classifier++ for metagenomic taxonomic classification-query evaluation.

Bioinformatics (Oxford, England)·2024
Same author

Efficient data acquisition and reconstruction for air-coupled ultrasonic robotic NDE.

Scientific reports·2024
Same author

Can machine learning predict late seizures after intracerebral hemorrhages? Evidence from real-world data.

Epilepsy & behavior : E&B·2024

Learn++ enables incremental learning without forgetting past knowledge or needing original data. This algorithm effectively handles new data, including unseen classes, for improved classifier performance.

Area of Science:

  • Machine Learning
  • Artificial Intelligence
  • Non-destructive Evaluation

Background:

  • Incremental learning algorithms are crucial for dynamic environments where data evolves over time.
  • Traditional classifiers often struggle with catastrophic forgetting when trained on new data batches.
  • Non-destructive evaluation (NDE) generates large, batched datasets, necessitating adaptable learning methods.

Purpose of the Study:

  • To introduce an incremental learning algorithm, Learn++, capable of acquiring new knowledge without forgetting prior information.
  • To develop a method that handles novel classes in incoming data without access to the original training dataset.
  • To apply and evaluate the algorithm in real-world scenarios like ultrasonic weld inspection.

Main Methods:

Related Experiment Videos

  • Learn++ utilizes an ensemble of classifiers, each trained on distinct subsets of available training data.
  • A weighted majority voting procedure combines the predictions of individual classifiers within the ensemble.
  • The algorithm is designed to be classifier-agnostic, compatible with various supervised learning techniques.
  • Confidence estimation is integrated into the decision-making process via the voting mechanism.

    • Main Results:

    • The Learn++ algorithm demonstrates effective incremental learning, retaining previously acquired knowledge.
    • It successfully incorporates new data, including previously unseen classes, without performance degradation.
    • Promising results were achieved in two distinct ultrasonic weld inspection applications, showcasing practical utility.
    • The ensemble approach and weighted voting contribute to synergistic generalization performance.

    Conclusions:

    • Learn++ provides a robust solution for incremental learning in scenarios with evolving datasets, such as NDE.
    • The algorithm's ability to learn without forgetting and handle new classes makes it highly adaptable.
    • Its independence from specific classifiers and integrated confidence estimation enhance its versatility and reliability.