Learning Bayesian Networks from Data with Unknown Labels: Theories and Experiments – This paper describes a technique for learning a probabilistic model for uncertain data. This model predicts some unknowns of an unknown sample. The prediction can be easily computed using a probability measure and also is accurate to be used as a tool for decision makers in a machine learning system. This probabilistic model has been used to classify data from multiple applications, and has been used for decision analysis and to assess the modelability of the model.

Recent results show that the Bayesian model of the data is able to capture the discriminative patterns of the data in a principled way. In this work, we take advantage of state-of-the-art Bayesian reasoning techniques to further reduce the Bayesian model complexity to some desirable levels. Firstly, it is shown to reduce the model complexity by a large margin to a few orders of magnitude for real-world data in an approach similar to the Bayes family.

Adversarial Data Analysis in Multi-label Classification

The Computational Phenomenon of Quantum-like Particles

Learning Bayesian Networks from Data with Unknown Labels: Theories and Experiments

Learning to rank for automatic speech synthesis

Multitask Learning with Class-level Generalized Linear ModelsRecent results show that the Bayesian model of the data is able to capture the discriminative patterns of the data in a principled way. In this work, we take advantage of state-of-the-art Bayesian reasoning techniques to further reduce the Bayesian model complexity to some desirable levels. Firstly, it is shown to reduce the model complexity by a large margin to a few orders of magnitude for real-world data in an approach similar to the Bayes family.