Foundations of Bayesian Networks
January 2009
in “
Elsevier eBooks
”
TLDR Bayesian networks are tools for modeling variables' probabilistic relationships, which can efficiently represent complex probabilities and help in making inferences.
The document from 2009 explains the theoretical foundations of Bayesian networks, which are graphical structures used to model probabilistic relationships among variables. It describes how these networks consist of a directed acyclic graph (DAG) and conditional probability distributions, and how they efficiently represent complex joint probability distributions with fewer values under the Markov condition. The document also discusses causality within Bayesian networks, using a study by Merck on finasteride as an example, where 1879 men participated, and the results showed significant hair regrowth in those treated with finasteride compared to a placebo. The U.S. FDA approved Propecia™ based on these findings. Additionally, the document covers inference in Bayesian networks using message-passing algorithms and software like HUGIN and Netica, which facilitate the computation of probabilities and handle continuous variables in Gaussian Bayesian networks.