| 摘要 |
PROBLEM TO BE SOLVED: To highly accurately identify an input pattern with even when the distributions of prototypes per class cannot be linearly separated and the prototypes have sparse distribution in which variances are different from each other. SOLUTION: When data of an input pattern is identified, data of preliminarily prepared prototypes is used to determine proximate prototypes per class of input pattern data, and a set of the determined proximate prototypes and near prototypes within a prescribed range is determined, and averages of proximate prototypes and near prototypes within the prescribed range are calculated per class, and distances weighted by averages of the input pattern and proximate prototypes per class are calculated, and a prototype of input pattern data is determined on the basis of calculated distances. COPYRIGHT: (C)2007,JPO&INPIT
|
