Generating attribute-based samples

1. A computer-implemented method, comprising:

training an encoder and a decoder of a latent variable model (LVM), utilizing labeled data and unlabeled data within a data set, wherein the LVM is selected from the group consisting of: a variational autoencoder (VAE), a Wasserstein autoencoder (WAE), an adversarial autoencoder (AAE), and a general adversarial network (GAN);

creating a latent space representation for the data set;

determining an explicit density model for the data set, utilizing the latent space representation for the data set;

determining a set of classifiers to identify which regions of the latent space representation are consistent with a predetermined set of labels;

sampling data points within the latent space representation for the data set that are consistent with a predetermined set of labels, utilizing rejection sampling; and

converting the sampled data points from a latent space representation to a data representation, utilizing the trained decoder.

2. The computer-implemented method of claim 1 , comprising training a classifier, utilizing the labeled data and associated labels within the data set; and generating new data having a predetermined set of labels, utilizing the LVM and the trained classifier.

3. The computer-implemented method of claim 1 , comprising training a classifier, utilizing the labeled data and associated labels within the data set; and generating new data having a predetermined set of labels, utilizing the LVM and the trained classifier, wherein generating the new data includes determining an explicit density model for the data set in the latent space representation.

4. The computer-implemented method of claim 2 , wherein the classifier is used to associate labels for the labeled data with the latent space representation of the data set.

5. The computer-implemented method of claim 2 , wherein generating the new data includes sampling data points within the latent space representation that are consistent with the predetermined set of labels.

6. The computer-implemented method of claim 5 , wherein the sampling is performed utilizing an explicit density model and the trained classifier.

7. The computer-implemented method of claim 5 , wherein the sampling includes rejection sampling that is used to sample points in the latent space representation that have the predetermined set of labels by identifying a region associated with the predetermined set of labels within the latent space representation.

8. The computer-implemented method of claim 7 , wherein Bayes' rule is applied during the rejection sampling.

9. A computer program product for generating attribute-based samples, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program instructions executable by a processor to cause the processor to perform a method comprising:

training, by the processor, an encoder and a decoder of a latent variable model (LVM), utilizing labeled data and unlabeled data within a data set, wherein the LVM is selected from the group consisting of: a variational autoencoder (VAE), a Wasserstein autoencoder (WAE), an adversarial autoencoder (AAE), and a general adversarial network (GAN);

creating, by the processor, a latent space representation for the data set;

determining, by the processor, an explicit density model for the data set, utilizing the latent space representation for the data set;

determining, by the processor, a set of classifiers to identify which regions of the latent space representation are consistent with a predetermined set of labels;

sampling, by the processor, data points within the latent space representation for the data set that are consistent with a predetermined set of labels, utilizing rejection sampling; and

converting, by the processor, the sampled data points from a latent space representation to a data representation, utilizing the trained decoder.

10. The computer program product of claim 9 , comprising:

training, by the processor, a classifier, utilizing the labeled data and associated labels within the data set, wherein the classifier is used to associate labels for the labeled data with the latent space representation of the data set.

11. The computer program product of claim 9 , comprising:

training, by the processor, a classifier, utilizing the labeled data and associated labels within the data set; and generating, by the processor, new data having a predetermined set of labels, utilizing the LVM and the trained classifier, wherein generating the new data includes sampling data points within the latent space representation that are consistent with the predetermined set of labels.

12. The computer program product of claim 11 , wherein the sampling is performed utilizing an explicit density model and the trained classifier.

13. The computer program product of claim 11 , wherein the sampling includes rejection sampling that is used to sample points in the latent space representation that have the predetermined set of labels by identifying a region associated with the predetermined set of labels within the latent space representation.

14. The computer program product of claim 13 , wherein Bayes' rule is applied during the rejection sampling.

15. A system, comprising:

a processor; and

logic integrated with the processor, executable by the processor, or integrated with and executable by the processor, the logic being configured to:

train an encoder and a decoder of a latent variable model (LVM), utilizing labeled data and unlabeled data within a data set, wherein the LVM is selected from the group consisting of: a variational autoencoder (VAE), a Wasserstein autoencoder (WAE), an adversarial autoencoder (AAE), and a general adversarial network (GAN);

create a latent space representation for the data set;

determine an explicit density model for the data set, utilizing the latent space representation for the data set;

determine a set of classifiers to identify which regions of the latent space representation are consistent with a predetermined set of labels;

sample data points within the latent space representation for the data set that are consistent with a predetermined set of labels, utilizing rejection sampling; and

convert the sampled data points from a latent space representation to a data representation, utilizing the trained decoder.

16. A computer-implemented method, comprising:

training an encoder and decoder of a latent variable model (LVM), utilizing labeled data and unlabeled data within a data set, wherein the LVM is selected from the group consisting of: a variational autoencoder (VAE), a Wasserstein autoencoder (WAE), an adversarial autoencoder (AAE), and a general adversarial network (GAN);

creating a latent space representation for the data set;

determining an explicit density model for the data set, utilizing the latent space representation for the data set;

determining a set of classifiers to identify which regions of the latent space representation are consistent with a predetermined set of labels;

sampling data points within the latent space representation for the data set that are consistent with the predetermined set of labels, utilizing rejection sampling; and

converting the sampled data points from a latent space representation to a data representation, utilizing the trained decoder.

17. A computer program product for generating attribute-based samples, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:

training, by the processor, an encoder and decoder of a latent variable model (LVM), utilizing labeled data and unlabeled data within a data set, wherein the LVM is selected from the group consisting of: a variational autoencoder (VAE), a Wasserstein autoencoder (WAE), an adversarial autoencoder (AAE), and a general adversarial network (GAN);

creating, by the processor, a latent space representation for the data set;

determining, by the processor, an explicit density model for the data set, utilizing the latent space representation for the data set;

determining, by the processor, a set of classifiers to identify which regions of the latent space representation are consistent with a predetermined set of labels;

sampling, by the processor, data points within the latent space representation for the data set that are consistent with the predetermined set of labels, utilizing rejection sampling; and

converting, by the processor, the sampled data points from a latent space representation to a data representation, utilizing a trained decoder of the LVM.