| 摘要 |
<P>PROBLEM TO BE SOLVED: To select an accurate classification score calculation unit from a plurality of classification score calculation units. <P>SOLUTION: The plurality of classification score calculation units calculate a classification score which is a posterior probability that prescribed input data belongs to each of a plurality of classes or a likelihood or a distance from a separation hyperplane and indicates that the probability of belonging to a corresponding class is higher when a value is larger, and store it in a classification score storage unit. All the classification scores are acquired from the classification score storage unit, and from all the classification scores, a relative certainty factor which indicates the probability that there is a correct class in a class or a class group corresponding to high-order N pieces of the classification scores and there is not a correct class in a class or a class group corresponding to high-order N pieces of the classification scores of the other classification score calculation units is calculated for each classification score calculation unit. The class or the class group to which input data belongs is determined on the basis of the classification score of the classification score calculation units of the highest relative certainty factor. <P>COPYRIGHT: (C)2013,JPO&INPIT |
