Consistent model selection of discrete Bayesian networks from incomplete data

Learning Bayesian Networks from Incomplete Data

Much of the current research in learning Bayesian Networks fails to eeectively deal with missing data. Most of the methods assume that the data is complete, or make the data complete using fairly ad-hoc methods; other methods do deal with missing data but learn only the conditional probabilities, assuming that the structure is known. We present a principled approach to learn both the Bayesian n...

Learning Discrete Bayesian Networks from Continuous Data

Real data often contains a mixture of discrete and continuous variables, but many Bayesian network structure learning and inference algorithms assume all random variables are discrete. Continuous variables are often discretized, but the choice of discretization policy has significant impact on the accuracy, speed, and interpretability of the resulting models. This paper introduces a principled ...

Estimating the Parameters of Mixed Bayesian Networks from Incomplete Data

Under complete data, there are closed-form maximum likelihood estimators for mixed Bayesian networks composed of discrete models [1], conditional Gaussian models [2] and conditional Gaussian regression models [2]. We describe an extension to Lauritzen’ expectation-maximisation (EM) algorithm [3], which estimates the parameters of discrete networks from incomplete data, to the more general case ...

Image Reconstruction from Incomplete Data

Model selection in incomplete data

Model selection in complete data is a common task for the applied researcher. However, in many scenarios data are incomplete which further complicates the task of model selection. In this talk, we will specify the problem of model selection in incomplete data and discuss several possible solutions using multiple imputation. First, we will define a new general measure for the correct model selec...