Learning device, learning method, and computer program product
According to an embodiment, a learning device includes a selecting unit, a learning unit, and an evaluating unit. The selecting unit performs a plurality of selection processes of selecting a plurality of groups including one or more learning samples from a learning sample storage unit, where respective learning samples are classified into any one of a plurality of categories. The learning unit learns a classification metric and obtains a set of a classification metric. The evaluating unit acquires two or more evaluation samples of different categories from an evaluation sample storage unit where respective evaluation samples are classified into any one of a plurality of categories; evaluates the classification metric included in the set of the classification metric using the two or more acquired evaluation samples; acquires a plurality of classification metric corresponding to the evaluation results from the set of the classification metric; and thereby generates an evaluation metric including the plurality of classification metric.
1. A learning device comprising:
a processor;
a selecting unit configured to perform a plurality of selection processes of selecting a plurality of groups including one or more learning samples from a learning sample storage unit that stores a plurality of learning samples in which respective learning samples are classified into any one of a plurality of categories;
a learning unit configured to learn a classification metric that classifies the plurality of groups for each of the plurality of selected groups and obtains a set of a classification metric; and
an evaluating unit configured to
acquire two or more evaluation samples of different categories from an evaluation sample storage unit that stores a plurality of evaluation samples in which respective evaluation samples are classified into any one of a plurality of categories,
evaluate the classification metric included in the set of the classification metric using the two or more acquired evaluation samples,
acquire a plurality of classification metric corresponding to the evaluation results from the set of the classification metric, and thereby
generate an evaluation metric including the plurality of classification metric.
2. The device according to claim 1 ,
wherein distribution of categories of the evaluation samples is different from distribution of categories of the learning samples, and
wherein the evaluating unit
evaluates a classification performance of the classification metric included in the set of the classification metric to classify evaluation samples of different categories using the two or more evaluation samples, and thereby
acquires a plurality of classification metric corresponding to the evaluated classification performance.
3. The device according to claim 1 , further comprising:
an input receiving unit configured to receive input of patterns including targets;
a feature amount calculating unit configured to calculate a feature amount of the patterns using the evaluation metric;
a similarity calculating unit configured to calculate a similarity between the feature amount and a standard feature amount serving as a recognition metric of the targets;
a recognition unit configured to recognize the targets using the similarity; and
an output control unit configured to cause an output unit to output the recognition results,
wherein the standard feature amount is a feature amount of the evaluation sample calculated using the evaluation metric.
4. The device according to claim 1 ,
wherein the evaluating unit evaluates classification metric included in the set of the classification metric using the similarity between evaluation samples of different categories.
5. The device according to claim 4 ,
wherein the evaluating unit
applies the two or more evaluation samples to a combination of an acquired classification metric and each of non-acquired classification metric,
specifies the maximum value of the similarities between the evaluation samples of different categories for each combination, and
acquires a non-acquired classification metric included in a combination in which the maximum value is smallest.
6. The device according to claim 1 ,
wherein the evaluating unit evaluates classification metric included in the set of the classification metric, by using at least one of an intra-class dispersion and an inter-class dispersion of the two or more evaluation samples.
7. The device according to claim 6 ,
wherein the evaluating unit
applies the two or more evaluation samples to respective classification metric included in the set of the classification metric,
calculates a value obtained by dividing the inter-class dispersion of the two or more evaluation samples by the intra-class dispersion for each classification metric, and
acquires a plurality of classification metric in the descending order of the calculated value.
8. The device according to claim 1 ,
wherein the learning unit learns the classification metric using a support vector machine, and
wherein the evaluating unit evaluates classification metric included in the set of the classification metric using a soft margin between evaluation samples of different categories.
9. The device according to claim 8 ,
wherein the evaluating unit
applies the two or more evaluation samples to a combination of an acquired classification metric and each of non-acquired classification metric,
specifies the maximum value of soft margins for each combination, and
acquires a non-acquired classification metric included in a combination in which the specified maximum value is smallest.
10. The device according to claim 4 ,
wherein the evaluating unit recursively acquires classification metric.
11. A learning method comprising:
performing a plurality of selection processes of selecting a plurality of groups including one or more learning samples from a learning sample storage unit that stores a plurality of learning samples in which respective learning samples are classified into any one of a plurality of categories;
learning a classification metric that classifies the plurality of groups for each of the plurality of selected groups and obtaining a set of a classification metric; and
evaluating that includes
acquiring two or more evaluation samples of different categories from an evaluation sample storage unit that stores a plurality of evaluation samples in which respective evaluation samples are classified into any one of a plurality of categories,
evaluating the classification metric included in the set of the classification metric using the two or more acquired evaluation samples,
acquiring a plurality of classification metric corresponding to the evaluation results from the set of the classification metric, and
generating an evaluation metric including the plurality of classification metric.
12. A computer program product comprising a non-transitory computer readable medium including programmed instructions, wherein the instructions, when executed by a computer, cause the computer to perform:
performing a plurality of selection processes of selecting a plurality of groups including one or more learning samples from a learning sample storage unit that stores a plurality of learning samples in which respective learning samples are classified into any one of a plurality of categories;
learning a classification metric that classifies the plurality of groups for each of the plurality of selected groups and obtaining a set of a classification metric; and
evaluating that includes
acquiring two or more evaluation samples of different categories from an evaluation sample storage unit that stores a plurality of evaluation samples in which respective evaluation samples are classified into any one of a plurality of categories,
evaluating the classification metric included in the set of the classification metric using the two or more acquired evaluation samples,
acquiring a plurality of classification metric corresponding to the evaluation results from the set of the classification metric, and
generating an evaluation metric including the plurality of classification metric.