Feature Discovery by Competitive Learning.
Abstract
This paper reports the results of our studies with an unsupervised learning paradigm which we have called 'Competitive Learning'. We have examined competitive learning using both computer simulation and formal analysis and have found that when it is applied to parallel networks of neuron-like elements, many potentially useful learning tasks can be accomplished. How a very simple competitive mechanism can be discovered a set of feature detectors which capture important aspects of the set of stimulus input patterns. How these feature detectors can form the basis of a multi-layer system that can serve to learn categorizations of stimulus sets which are not linearly separable. How the use of correlated stimuli can be served as a kind of 'teaching' input to the system to allow the development of feature detectors which would not develop otherwise. Competitive learning is an essentially non-associative statistical learning scheme. We certainly imagine that other kinds of learning mechanisms will be involved in the building of associations among patterns of activation in a more complete neural network.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1984
- Accession Number
- ADA145052
Entities
People
- D. E. Rumelhart
- D. Zipser
Organizations
- University of California, San Diego