Title

Supervised Competitive Learning

Authors

Thomas H. Fuller Jr., Washington University in St Louis
Takayuki D. Kimura, Washington University in St Louis

Document Type

Technical Report

Publication Date

1993

Filename

WUCS-93-45.pdf

DOI:

10.7936/K7707ZRH

Technical Report Number

WUCS-93-45

Abstract

Supervised Competitive Learning (SCL) assembles a set of learning modules into a supervised learning system to address the stability-plasticity dilemma. Each learning module acts as a similarity detector for a prototype, and includes prototype resetting (akin to that of the ART) to respond to new prototypes. SCL has usually employed backpropagation networks as the learning modules. It has been tested with two feature abstractors: about 30 energy-based features, and a combination of energy-based and graphical features (about 60). Anout 75 subjects have been involved. In recent testing (15 college students), SCL recognized 99% (energy features only) of test digits, 91% (energy) and 96.6% (energy/graphical) of test letters, and 85% of test gestures (energy/graphical)/ SCL has also been tested with fuzzy sets as learning modules for recognizing handwriting digits and handwritten gestures, recognizing 97% of test digits, and 91% of test gestures.

Comments

Permanent URL: http://dx.doi.org/10.7936/K7707ZRH

Recommended Citation

Fuller, Thomas H. Jr. and Kimura, Takayuki D., "Supervised Competitive Learning" Report Number: WUCS-93-45 (1993). All Computer Science and Engineering Research.
https://openscholarship.wustl.edu/cse_research/540