Learning method, device, learning apparatus and program

1. A learning method comprising the steps of:

acquiring, by a device, a plurality of samples;

dividing, by the device, the plurality of samples into a plurality of clusters;

extracting, by the device, samples from each of the plurality of clusters according to an effectiveness of each of the plurality of clusters received from a learning apparatus, wherein a number of samples extracted from the each of the plurality of clusters is proportional to the effectiveness of the each of the plurality of clusters;

transmitting, by the device, the extracted samples to the learning apparatus;

learning, by the learning apparatus, the extracted samples;

calculating, by the learning apparatus, for each of the plurality of clusters, an effectiveness in learning samples of the extracted samples belonging to a cluster of the plurality of clusters from learning results; and

transmitting, by the learning apparatus, the effectiveness of each of the plurality of clusters to the device.

2. The learning method according to claim 1 , wherein

any one label out of a plurality of types of labels is assigned to each of the extracted samples, and

the step of calculating includes the steps of:

counting, for each of the plurality of types of labels assigned to the extracted samples, the number of samples where the label is assigned, and

calculating, for each of the plurality of clusters, the effectiveness of the cluster to be larger when the number of samples having the label assigned to the samples belonging to the cluster is small.

3. The learning method according to claim 1 , wherein the step of learning includes the steps of:

calculating, for the extracted samples, probabilities corresponding to each of the plurality of types of labels using a learning model generated by learning the extracted samples, and

calculating, for each of the plurality of clusters, the effectiveness of the cluster to be larger when a degree of variation of the probabilities corresponding to each of the plurality of labels for the samples belonging to the cluster is small.

4. The learning method according to claim 1 , wherein

in the step of extracting of the samples, samples farther from a centroid of the cluster have higher probability of being extracted from the cluster.

5. The learning method according to claim 1 , wherein

a plurality of cycles having from the step of-acquiring of the plurality of samples to the step of-transmitting of the effectiveness for each of the plurality of clusters to the device as one cycle are repeated,

the effectiveness of each of the plurality of clusters received from the learning apparatus is an effectiveness of each of a plurality of clusters in a previous cycle, and

the step of extracting of the sample includes the steps of:

associating the plurality of clusters in the previous cycle with a corresponding plurality of clusters identified in a current cycle based on centroids of clusters, and

extracting samples from each of the plurality of clusters according to the effectiveness of a cluster of the plurality of clusters in the previous cycle corresponding to a cluster of the plurality of clusters identified in the current cycle.

6. The learning method according to claim 2 , wherein in the step of extracting of the samples, samples farther from a centroid of the cluster have higher probability of being extracted from the cluster.

7. The learning method according to claim 3 , wherein in the step of extracting of the samples, samples farther from a centroid of the cluster have higher probability of being extracted from the cluster.

8. The learning method according to claim 2 , wherein

a plurality of cycles having from the step of acquiring of the plurality of samples to the step of transmitting of the effectiveness for each of the plurality of clusters to the device as one cycle are repeated,

the effectiveness of each of the plurality of clusters received from the learning apparatus is an effectiveness of each of a plurality of clusters in a previous cycle, and

the step of extracting of the sample includes the steps of:

associating the plurality of clusters in the previous cycle with a corresponding plurality of clusters identified in a current cycle based on centroids of clusters, and

extracting samples from each of the plurality of clusters according to the effectiveness of a cluster of the plurality of clusters in the previous cycle corresponding to a cluster of the plurality of clusters identified in the current cycle.

9. The learning method according to claim 3 , wherein

a plurality of cycles having from the step of acquiring of the plurality of samples to the step of transmitting of the effectiveness for each of the plurality of clusters to the device as one cycle are repeated,

the effectiveness of each of the plurality of clusters received from the learning apparatus is an effectiveness of each of a plurality of clusters in a previous cycle, and

the step of extracting of the sample includes the steps of:

associating the plurality of clusters in the previous cycle with a corresponding plurality of clusters identified in a current cycle based on centroids of clusters, and

extracting samples from each of the plurality of clusters according to the effectiveness of a cluster of the plurality of clusters in the previous cycle corresponding to a cluster of the plurality of clusters identified in the current cycle.

10. The learning method according to claim 4 , wherein

a plurality of cycles having from the step of acquiring of the plurality of samples to the step of transmitting of the effectiveness for each of the plurality of clusters to the device as one cycle are repeated,

the effectiveness of each of the plurality of clusters received from the learning apparatus is an effectiveness of each of a plurality of clusters in a previous cycle, and

the step of extracting of the sample includes the steps of:

associating the plurality of clusters in the previous cycle with a corresponding plurality of clusters identified in a current cycle based on centroids of clusters, and

extracting samples from each of the plurality of clusters according to the effectiveness of a cluster of the plurality of clusters in the previous cycle corresponding to a cluster of the plurality of clusters identified in the current cycle.

11. A device comprising:

a processor; and

a memory device storing instructions that, when executed by the processor, configure the processor to:

acquire a plurality of samples;

divide the plurality of samples into a plurality of clusters;

extract samples from each of the plurality of clusters according to a ratio of an effectiveness of each of the plurality of clusters received from a learning apparatus, wherein a number of samples extracted from the each of the plurality of clusters is proportional to the effectiveness of the each of the plurality of clusters; and

transmit the extracted samples to the learning apparatus.

12. A learning apparatus comprising:

a processor; and

a memory device storing instructions that, when executed by the processor, configure the processor to:

learn samples extracted by the device of claim 11 ;

calculate, per clusters basis, the plurality of clusters being obtained by clustering the samples by the device, an effectiveness in learning the samples belonging to a cluster of the clusters from learning results; and

transmit the effectiveness of each of the clusters to the device.

13. A non-transitory computer readable medium storing a program, wherein executing of the program causes a computer to operate as the device according to claim 11 .

14. A non-transitory computer readable medium storing a program, wherein executing of the program causes a computer to operate as the learning apparatus according to claim 12 .