Classification system and learning system
A classification system according to an embodiment includes a score calculation unit, a determination unit, and a classification unit. The score calculation unit calculates respective scores of predetermined classes from input data. The determination unit determines whether the input data belongs to anyone of the classes based on the respective scores of the classes, which are calculated by the score calculation unit. The classification unit determines which one of the classes the input data belongs to, based on the calculated scores when the determination unit determines that the input data belongs to anyone of the classes and determines that the input data belongs to an unknown class that is other than the classes when the determination unit determines that the input data does not belong the classes.
1. A classification system comprising:
a score calculation unit configured to calculate respective scores of predetermined classes from input data;
a determination unit configured to determine whether the input data belongs to anyone of the classes, based on the respective scores of the classes, which are calculated by the score calculation unit; and
a classification unit configured to determine to which one of the classes the input data belongs, based on the respective scores when the determination unit determines that the input data belongs to anyone of the classes, and to determine that the input data belongs to an unknown class that is other than the classes when the determination unit determines that the input data does not belong to the classes,
wherein the determination unit includes:
trained determination models, corresponding to respective classes, configured to determine whether a vector including the respective scores of the classes belongs to a corresponding class; and
a selection unit configured to supply a vector including the respective scores, which are calculated from the input data, to one of the determination models corresponding to the class with the largest score among the respective scores of the classes, which are calculated by the score calculation unit, and
wherein a determination result by the one of the determination models, to which the vector is supplied, is supplied to the classification unit.
2. The classification system according to claim 1 , wherein the score calculation unit comprises:
a trained calculation model configured to calculate a first value which represents a respective certainty of each of the classes from the input data, and
a conversion unit configured to convert the first value of each of the classes calculated by the calculation model into a second value within a predetermined range by using a sigmoid function and outputs, as each of the calculated scores, each second value into which the first value of each of the classes is converted.
3. The classification system according to claim 1 , wherein the determination unit determines whether the input data belongs to one of the classes, based on the largest score among the respective scores of the classes, which are calculated by the score calculation unit.
4. The classification system according to claim 1 , wherein the determination unit includes one-class support vector machines as the determination models.
5. The classification system according to claim 1 ,
wherein each one of the determination models that correspond to the respective classes is obtained through training using a set of vectors,
wherein each vector included in the set of vectors includes, as elements, respective scores to the classes, which are calculated by the score calculation unit, and
wherein each vector included in the set of vectors has, as one of the elements, a score corresponding to a class of the one of the determination models, and the score is higher than the other scores.
6. A non-transitory computer-readable medium storing a program for causing a computer provided in a classification system to function as:
a score calculation unit configured to calculate respective scores of predetermined classes from input data;
a determination unit configured to determine whether the input data belongs to anyone of the classes, based on the respective scores of the classes, which are calculated by the score calculation unit; and
a classification unit configured to determine to which one of the classes the input data belongs, based on the respective scores when the determination unit determines that the input data belongs to anyone of the classes, and to determine that the input data belongs to an unknown class that is other than the classes when the determination unit determines that the input data does not belong to the classes,
wherein the determination unit includes:
trained determination models, corresponding to respective classes, configured to determine whether a vector including the respective scores of the classes belongs to a corresponding class; and
a selection unit configured to supply a vector including the respective scores, which are calculated from the input data, to one of the determination models corresponding to the class with the largest score among the respective scores of the classes, which are calculated by the score calculation unit, and
wherein a determination result by the one of the determination models, to which the vector is supplied, is supplied to the classification unit.