Abstract: Generative models have recently become an effective approach for anomaly detection by leveraging auto-encoders to model high-dimensional data and identify anomalies ... the frameworks of ...

To discover features and their conditional independence structure, we develop pimaDAG, a variational autoencoder framework that learns ... BN on simultaneously discovered key features in a fully ...

They often require a large amount of training data ... and a text to vec component, and introduces an improved hierarchical speech synthesizer built on the hierarchical conditional VAE or Variational ...

Generating synthetic data is useful when you have imbalanced training data for a particular class, for example, generating synthetic females in a dataset of employees that has many males but few ...

Center for Bio/Molecular Science & Engineering (Code 6900), U.S. Naval Research Laboratory, Washington, D.C. 20375, United States Article Views are the COUNTER-compliant sum of full text article ...