Title page -- Contents -- Foreword -- Introduction -- Preliminaries -- Random variables and conditional independence -- Graph theory -- Markov properties -- The Markov blanket -- Equivalence of DAGs -- Bayesian networks -- Bayesian network specification -- Bayesian parameter specification -- Learning Bayesian Networks from Data -- The basics -- Learning parameters -- The Maximum-Likelihood approach -- The Bayesian approach -- Learning models -- The penalised likelihood approach -- The Bayesian approach -- Learning via the marginal likelihood -- Determining the hyper parameter -- Marginal and penalised likelihood -- Search methodologies -- Model search space -- Traversal strategy -- Monte Carlo Methods and MCMC Simulation -- Monte Carlo methods -- Importance sampling -- Choice of the sampling distribution -- Markov chain Monte Carlo-MCMC -- Markov chains -- The invariant target distribution -- Reaching the invariant distribution -- Metropolis-Hastings sampling -- Gibbs sampling -- Mixing, burn-in and convergence of MCMC -- The importance of blocking -- Learning models via MCMC -- Sampling models -- Sampling edges -- Blocking edges -- Blocks and Markov blankets -- Sampling blocks -- Validity of the sampler -- The MB-MCMC model sampler -- Evaluation -- Conclusion -- Learning from Incomplete Data -- The concept of incomplete data -- Missing data mechanisms -- Learning from incomplete data -- Likelihood decomposition -- Complications for learning parameters -- Bayesian sequential updating -- Complications for learning models -- Principled iterative methods -- Expectation Maximisation-EM -- Structural EM-SEM -- Data Augmentation-DA -- DA and eliminating the P-step-DA-P -- DA-P and model learning-MDA-P -- Efficiency issues of MDA-P -- Properties of the sub-MCMC samplers -- Interdependence between samplers -- Imputation via importance sampling.

The general idea -- Importance sampling in the I-step-ISMDA-P -- Generating new population vs. re-weighing -- The marginal likelihood as predictive distribution -- The eMC4 sampler -- Evaluation-proof of concept -- Conclusion -- Ad-hoc and heuristic methods -- Available cases analysis -- Bound and Collapse-BC -- Markov Blanket Predictor-MBP -- The general idea -- Approximate predictive distributions -- Parameter estimation -- Prediction and missing parents -- Predictive quality -- Selecting predictive variables -- Implementation of MBP -- Parameter estimation -- Model learning -- Conclusion and discussion -- Conclusion -- References.