摘要 |
This invention relates generally to methods and apparatus for pattern classification, and, in particular, to pattern classification methods and apparatus which distribute the classification criteria across an artificial neural network. The classification criteria for a pattern class are stored distributively in the neural network in two aspects. First, they manifest themselves as one or more (quantum) levels of templates, each of which represents a fuzzily unique perspective of the pattern class. Second, the templates at each level are represented by a set of fuzzy windows, each of which defines a classification criterion of a corresponding feature of the pattern class. The resultant artificial neural network adapts to produce a "match metric" which indicates the degree of similarity between a new input pattern and a pattern stored distributively in the neural network in the form of the plurality of fuzzy windows.
|