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

Robert Peharz1, Franz Pernkopf

  • 1Signal Processing and Speech Communication Laboratory, Graz University of Technology, Austria.

Neurocomputing
|April 17, 2012
PubMed
Summary
This summary is machine-generated.

This study introduces a new framework for approximate nonnegative matrix factorization (NMF) that constrains the L1 norm for sparser data representations. The proposed methods demonstrate competitive or superior performance compared to existing NMF approaches.

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

Adaptive Variational Inference: Beyond Bethe, Tree-Reweighted, and Convex Free Energies.

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

Acoustic COVID-19 Detection Using Multiple Instance Learning.

IEEE journal of biomedical and health informatics·2024
Same author

Dataset of directional room impulse responses for realistic speech data.

Data in brief·2024
Same author

Self-Guided Belief Propagation - A Homotopy Continuation Method.

IEEE transactions on pattern analysis and machine intelligence·2022
Same author

Lung Sound Classification Using Co-Tuning and Stochastic Normalization.

IEEE transactions on bio-medical engineering·2022
Same author

Crackle Detection In Lung Sounds Using Transfer Learning And Multi-Input Convolutional Neural Networks.

Annual International Conference of the IEEE Engineering in Medicine and Biology Society. IEEE Engineering in Medicine and Biology Society. Annual International Conference·2021
Same journal

Automated Behavior Analysis in the Novel Object Recognition Test.

Neurocomputing·2026
Same journal

CrunchLLM: Multitask LLMs for Structured Business Reasoning and Outcome Prediction.

Neurocomputing·2026
Same journal

Deep Learning for analyzing chaotic dynamics in biological time series: Insights from frog heart signals.

Neurocomputing·2026
Same journal

SymRefine: A symbolic regression approach for refining and compressing neural networks.

Neurocomputing·2026
Same journal

Artificial intelligence without restriction surpassing human intelligence with probability one: Theoretical insight into secrets of the brain with AI twins of the brain.

Neurocomputing·2025
Same journal

ShaderNN: A Lightweight and Efficient Inference Engine for Real-time Applications on Mobile GPUs.

Neurocomputing·2025
See all related articles

Area of Science:

  • Machine Learning
  • Data Science
  • Linear Algebra

Background:

  • Nonnegative matrix factorization (NMF) aims for sparse, part-based data representations, but this is not always guaranteed.
  • Existing NMF methods often enforce sparsity by constraining the L2 norm of factor matrices.
  • Limited research has explored using the L1 norm, a more natural measure of sparsity, in NMF.

Purpose of the Study:

  • To propose a novel framework for approximate nonnegative matrix factorization (NMF).
  • To incorporate constraints on the L1 norm of either the basis or coefficient matrix.
  • To enhance the inherent sparsity and part-based representation capabilities of NMF.

Main Methods:

  • Developed a flexible framework for approximate NMF incorporating L1 norm constraints.

Related Experiment Videos

  • Integrated existing unconstrained NMF techniques, such as multiplicative update rules and alternating nonnegative least-squares (ANLS).
  • Applied the framework to enforce sparsity on the basis matrix or the coefficient matrix.

    • Main Results:

    • Experimental results demonstrate the effectiveness of the proposed L1-norm constrained NMF framework.
    • The new methods achieve sparsity levels comparable to or better than existing approaches.
    • The framework offers improved or competitive performance in NMF tasks.

    Conclusions:

    • The proposed framework effectively enhances sparsity in NMF by constraining the L1 norm.
    • This approach provides a versatile method for achieving desired sparse representations in nonnegative data.
    • The L1-norm constraint offers a valuable alternative for improving NMF performance and interpretability.