Parameter estimation

Part of the Information Science and Statistics book series (ISS)


Assume that you know the structure of a Bayesian network model over the variables \( \mathcal{U} \), but you do not have any estimates for the conditional probabilities. On the other hand, you have access to a database of cases, i.e., a set of simultaneous values for some of the variables in \( \mathcal{U} \). You can now use these cases to estimate the parameters of the model, namely the conditional probabilities. In this chapter we consider two approaches for handling this problem: First we show how a database of cases can be used to estimate the parameters once and for all (so-called batch learning). After that we shall investigate the situation where the cases are accumulated sequentially, i.e., we would like to adapt the model as each new case arrives. The reader is expected to be familiar with Section 1.5.


Maximum Likelihood Estimate Bayesian Network Prior Distribution Fading Factor Dirichlet Distribution 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer Science +Business Media, LLC 2007

Personalised recommendations