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

Intrinsic generalization analysis of low dimensional representations.

Xiuwen Liu1, Anuj Srivastava, DeLiang Wang

  • 1Department of Computer Science, Florida State University, Palmetto St. LOVE Building Rm 250, Tallahassee, FL 32306-4530, USA. liux@cs.fsu.edu

Neural Networks : the Official Journal of the International Neural Network Society
|July 10, 2003
PubMed
Summary

This study introduces intrinsic generalization as a new metric for image representation. Better intrinsic generalization, achieved through a novel nonlinear projection, leads to improved image recognition performance.

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

A speech prediction model based on codec modeling and transformer decoding.

Computer speech & language·2026
Same author

STITCH links cellular morphology and gene expression in spatial transcriptomics.

bioRxiv : the preprint server for biology·2026
Same author

A Molecular Trimming Strategy for Hypoxia-Tolerant Photosensitizers With Enhanced cGAS-STING Activation.

Angewandte Chemie (International ed. in English)·2026
Same author

Towards decoupling frontend enhancement and backend recognition in monaural robust ASR.

Computer speech & language·2026
Same author

Efficacy of SWIM technology combined with direct aspiration first pass technique for large vessel occlusion in acute ischemic stroke.

American journal of translational research·2026
Same author

Comparative efficacy of levosimendan and dobutamine in sepsis-related cardiac impairment: a meta-analysis.

European journal of clinical pharmacology·2026

Area of Science:

  • Computer Vision
  • Machine Learning
  • Image Processing

Background:

  • Low-dimensional image representations define equivalence classes, termed intrinsic generalization.
  • Intrinsic generalization offers a more fundamental measure of representation quality than traditional recognition performance.
  • Recognition performance is often confounded by dataset specifics.

Purpose of the Study:

  • To evaluate the limitations of linear subspace image representations.
  • To propose and validate a novel nonlinear representation for improved intrinsic generalization.
  • To demonstrate the correlation between intrinsic generalization and recognition accuracy.

Main Methods:

  • Sampling intrinsic generalization to analyze linear subspace limitations.

Related Experiment Videos

  • Developing a nonlinear projection method utilizing marginal densities of filter responses.
  • Applying linear projections to the marginals for the final representation.

    • Main Results:

    • Linear subspace representations exhibit limitations in intrinsic generalization.
    • The proposed nonlinear representation demonstrates superior intrinsic generalization.
    • Improved intrinsic generalization directly correlates with enhanced recognition performance on large datasets.

    Conclusions:

    • Intrinsic generalization provides a robust metric for evaluating image representations.
    • Nonlinear representations can overcome the limitations of linear methods.
    • The proposed method offers a pathway to more effective image recognition systems.