Semiparametric Bayesian networks

We introduce semiparametric Bayesian networks that combine parametric and nonparametric conditional probability distributions. Their aim is to incorporate the advantages of both components: bounded complexity models flexibility ones. demonstrate generalize two well-known types networks: Gaussian kernel density estimation networks. For this purpose, we consider different distributions required in a network. In addition, present modifications algorithms (greedy hill-climbing PC) learn structure network from data. To realize this, employ score function based on cross-validation. using validation dataset, apply an early-stopping criterion avoid overfitting. evaluate applicability proposed algorithm, conduct exhaustive experiment synthetic data sampled by mixing linear nonlinear functions, multivariate normal networks, real UCI repository, bearings degradation As result experiment, conclude algorithm accurately learns combination components, while achieving performance comparable with those provided state-of-the-art methods.