Learning Probabilistic Model Parameters
Richard E. Neapolitan, Xia Jiang · 2018
This chapter discusses learning discrete parameters. Neapolitan (2004) shows a method for learning the parameters in a Gaussian Bayesian network. In a Bayesian network the DAG is called the structure. The chapter considers datasets in which every value of every variable is recorded in every case. It shows how to update parameters based on data containing data items missing at random. The method for learning parameters in a Bayesian network follows readily from the method for learning a single parameter. Eliciting Bayesian networks from experts can be a laborious and difficult process in the case of large networks. In a Bayesian network, the conditional probability distributions are called the parameters. The chapter concerns the probability of heads using a beta distribution with parameters a and b . It illustrates the method with binomial variables.