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

Computational neural networks: enhancing supervised learning algorithms via self-organization.

R M Holdaway1, M W White

  • 1Department of Electrical and Computer Engineering, North Carolina State University, Raleigh, NC 27695-7911.

International Journal of Bio-Medical Computing
|April 1, 1990
PubMed
Summary
This summary is machine-generated.

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

Toxoplasma co-opts host gene expression by injection of a polymorphic kinase homologue.

Nature·2006
Same author

Defining the cell cycle for the tachyzoite stage of Toxoplasma gondii.

Molecular and biochemical parasitology·2001
Same author

The DNA sequence of the simian varicella virus genome.

Virology·2001
Same author

Toxoplasma gondii: characterization of temperature-sensitive tachyzoite cell cycle mutants.

Experimental parasitology·2001
Same author

Abdominal perfusion pressure: a superior parameter in the assessment of intra-abdominal hypertension.

The Journal of trauma·2000
Same author

Renewal-process approximation of a stochastic threshold model for electrical neural stimulation.

Journal of computational neuroscience·2000
Same journal

Commentary on a futuristic model of patient record systems and telemedicine.

International journal of bio-medical computing·1996
Same journal

Nonlinear eye movement detection method for drowsiness studies.

International journal of bio-medical computing·1996
Same journal

Segmentation of auditory brainstem response signals.

International journal of bio-medical computing·1996
Same journal

A comparison of neural network and Bayes recognition approaches in the evaluation of the brainstem trigeminal evoked potentials in multiple sclerosis.

International journal of bio-medical computing·1996
Same journal

Methodology for using the UMLS as a background knowledge for the description of surgical procedures.

International journal of bio-medical computing·1996
Same journal

An MLP-based model for identifying qEEG in depression.

International journal of bio-medical computing·1996
See all related articles

A novel neural network design using a Kohonen feature map front end improves pattern recognition, especially for disjoint decision regions. This architecture enhances classifier performance by pre-sensitizing network units to input features during self-organization.

Area of Science:

  • Artificial Intelligence
  • Machine Learning
  • Neural Networks

Background:

  • Conventional feedforward neural networks face challenges with disjoint decision regions in pattern recognition tasks.
  • Feature extraction and representation are critical for effective classification.

Purpose of the Study:

  • To propose and evaluate a novel neural network processing scheme combining a self-organizing Kohonen feature map with a feedforward classifier.
  • To investigate the performance improvements of this architecture compared to conventional methods.

Main Methods:

  • A hybrid neural network architecture was developed, integrating a Kohonen feature map (KFM) as a preprocessing layer.
  • Benchmarking studies were conducted using artificial statistical pattern recognition tasks.
  • Performance was evaluated based on classification accuracy, particularly for datasets with disjoint decision regions.

Related Experiment Videos

Main Results:

  • The proposed architecture demonstrated significantly superior performance over conventional feedforward networks when dealing with disjoint decision regions.
  • The self-organization process of the KFM enabled internal units in the classifier to become feature-specific early in training.
  • This pre-sensitization led to more efficient and accurate classification.

Conclusions:

  • The integration of a self-organizing Kohonen feature map as a front end to feedforward classifiers offers a significant advantage for pattern recognition tasks with disjoint decision regions.
  • This approach enhances classification accuracy by leveraging the feature-learning capabilities of self-organizing maps.
  • The proposed scheme represents a promising advancement in neural network design for complex pattern recognition problems.