Clustering device, clustering method, and computer program product
According to one embodiment, a communication device includes one or more processors. The processors decide on a first-type parameter representing a conversion operation for converting input data into first-type conversion data. The processors calculate first-type predicted distributions based on second-type parameters, each representing one of the plurality of clusters. The processors update the first-type parameter and the second-type parameters so as to achieve optimization of first-type differences representing differences between the first-type predicted distributions and a target distribution, and second-type differences representing differences between the first-type predicted distributions and second-type predicted distributions that indicate probability at which second-type converted data. The second-type converted data is obtained by converting data, which is formed by augmentation of the input data, using the first-type parameter belongs to the clusters. The processors classify the input data into the plurality of clusters based on the first-type predicted distributions calculated using the updated second-type parameters.
1. A clustering device comprising:
one or more processors configured to:
decide on a first-type parameter representing a parameter of a conversion operation for converting input data into first-type conversion data, the input data being subjected to clustering for classifying data into a plurality of clusters;
based on a plurality of second-type parameters, each representing a parameter of one of the plurality of clusters, calculate a plurality of first-type predicted distributions indicating probability at which the first-type conversion data belongs to the clusters;
update the first-type parameter and the second-type parameters so as to achieve optimization of first-type differences representing differences between the first-type predicted distributions and a target distribution, and second-type differences representing differences between the first-type predicted distributions and second-type predicted distributions, the second-type predicted distributions indicating probability at which second-type converted data, the second-type converted data being obtained by converting data, which is formed by augmentation of the input data, using the first-type parameter belongs to the clusters;
and classify the input data into the plurality of clusters based on the first-type predicted distributions calculated using the updated second-type parameters.
2. The clustering device according to claim 1 , wherein the second-type predicted distributions indicate probability at which the second-type converted data that is obtained by converting data, which is formed by adding perturbation to the input data, using the first-type parameter belongs to the clusters.
3. The clustering device according to claim 2 , wherein the second-type predicted distributions indicate probability at which the second-type conversion data that is obtained by converting data, which is obtained by adding perturbation to the input data in direction in which amount of change is greater than other directions, using the first-type parameter belongs to the cluster.
4. The clustering device according to claim 1 , wherein the one or more processors update the first-type parameter and the second-type parameters so as to achieve optimization of weighted sum of the first-type differences and the second-type differences.
5. The clustering device according to claim 1 , wherein the one or more processors decide on the first-type parameter using an autoencoder.
6. The clustering device according to claim 1 , wherein the conversion operation is performed using a neural network.
7. The clustering device according to claim 1 , wherein the one or more processors calculate initial value of the second-type parameters by performing clustering of the first-type conversion data.
8. The clustering device according to claim 1 , wherein the one or more processors calculate the target distribution using a value indicating a power of the first-type predicted distributions.
9. The clustering device according to claim 8 , wherein the one or more processors calculate the target distribution using a value indicating s-th power of the first-type predicted distributions, s being a real value equal to greater than 1.
10. The clustering device according to claim 1 , wherein the one or more processors calculate the target distribution using sum of the plurality of first-type predicted distributions corresponding to the plurality of clusters.
11. A clustering method comprising:
deciding on a first-type parameter representing a parameter of a conversion operation for converting input data into first-type conversion data, the input data being subjected to clustering for classifying data into a plurality of clusters;
based on a plurality of second-type parameters, each representing a parameter of one of the plurality of clusters, calculating a plurality of first-type predicted distributions indicating probability at which the first-type conversion data belongs to the clusters;
updating the first-type parameter and the second-type parameters so as to achieve optimization of first-type differences representing differences between the first-type predicted distributions and a target distribution, and second-type differences representing differences between the first-type predicted distributions and second-type predicted distributions, the second-type predicted distributions indicating probability at which second-type converted data, the second-type converted data being obtained by converting data, which is formed by augmentation of the input data, using the first-type parameter belongs to the clusters;
and classifying the input data into the plurality of clusters based on the first-type predicted distributions calculated using the updated second-type parameters.
12. A computer program product having a non-transitory computer readable medium including programmed instructions,
wherein the instructions, when executed by a computer, cause the computer to perform: deciding on a first-type parameter representing a parameter of a conversion operation for converting input data into first-type conversion data, the input data being subjected to clustering for classifying data into a plurality of clusters;
based on a plurality of second-type parameters, each representing a parameter of one of the plurality of clusters, calculating a plurality of first-type predicted distributions indicating probability at which the first-type conversion data belongs to the clusters; updating the first-type parameter and the second-type parameters so as to achieve optimization of first-type differences representing differences between the first-type predicted distributions and a target distribution, and second-type differences representing differences between the first-type predicted distributions and second-type predicted distributions, the second-type predicted distributions indicating probability at which second-type converted data, the second-type converted data being obtained by converting data, which is formed by augmentation of the input data, using the first-type parameter;
and classifying the input data into the plurality of clusters based on the first-type predicted distributions calculated using the updated second-type parameters.