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

Classification ability of single hidden layer feedforward neural networks.

G B Huang1, Y Q Chen, H A Babri

  • 1School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore.

IEEE Transactions on Neural Networks
|February 6, 2008
PubMed
Summary

Single hidden layer feedforward neural networks (SLFNs) can create complex, arbitrary-shaped decision regions. This capability extends beyond perceptrons to various activation functions, proving their versatility in multidimensional pattern recognition.

Related Concept Videos

Classification of Systems-I01:26

Classification of Systems-I

Linearity is a system property characterized by a direct input-output relationship, combining homogeneity and additivity.
Homogeneity dictates that if an input x(t) is multiplied by a constant c, the output y(t) is multiplied by the same constant. Mathematically, this is expressed as:

You might also read

Related Articles

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

Sort by
Same author

Test of CP Symmetry in the Neutral Decays of Λ via J/ψ→ΛΛ[over ¯].

Physical review letters·2026
Same author

[Association between blood selenium level and depressive symptoms in older adults aged 65 years and above in 18 longevity areas in China].

Zhonghua liu xing bing xue za zhi = Zhonghua liuxingbingxue zazhi·2026
Same author

[Association between urinary nickel levels and impaired kidney function in older adults aged 65 years and above in 18 longevity areas in China].

Zhonghua liu xing bing xue za zhi = Zhonghua liuxingbingxue zazhi·2026
Same author

[Associations between dietary patterns and colorectal neoplasia in Chinese adults].

Zhonghua liu xing bing xue za zhi = Zhonghua liuxingbingxue zazhi·2026
Same author

[Analysis of the detection of occupational contraindications in Guangzhou area from 2021 to 2023].

Zhonghua lao dong wei sheng zhi ye bing za zhi = Zhonghua laodong weisheng zhiyebing zazhi = Chinese journal of industrial hygiene and occupational diseases·2026
Same author

Precise Measurement of the Chromoelectric Dipole Moment of the Charm Quark.

Physical review letters·2026

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Neural Networks

Background:

  • Multilayer perceptrons with hard-limiting activation functions form complex decision regions.
  • Two-layer perceptrons (one hidden layer) create single convex decision regions.
  • Three-layer perceptrons (two hidden layers) form arbitrary disjoint decision regions.

Purpose of the Study:

  • To prove that single hidden layer feedforward neural networks (SLFNs) can form arbitrary-shaped disjoint decision regions.
  • To extend this finding beyond perceptrons to a wider range of activation functions.
  • To demonstrate the capability of SLFNs in multidimensional cases.

Main Methods:

  • Theoretical analysis of single hidden layer feedforward neural networks (SLFNs).

Related Experiment Videos

  • Examination of various activation functions including continuous bounded, bounded nonconstant, and unbounded types.
  • Proof of capability to form arbitrary disjoint decision regions in multidimensional spaces.

    • Main Results:

    • Single hidden layer feedforward neural networks (SLFNs) with continuous bounded nonconstant activation functions can form arbitrary disjoint decision regions.
    • SLFNs with arbitrary bounded activation functions (continuous or not) having unequal limits at infinities also form arbitrary disjoint decision regions.
    • SLFNs utilizing some unbounded activation functions can similarly generate arbitrary-shaped disjoint decision regions.

    Conclusions:

    • Single hidden layer feedforward neural networks (SLFNs) possess greater flexibility in forming decision regions than previously established.
    • The choice of activation function significantly influences the complexity and shape of decision regions achievable by SLFNs.
    • These findings broaden the understanding of SLFNs' representational power in multidimensional pattern classification.