Parameter Estimation with Trees

Overview

In BayesiaLab, a Conditional Probability Distribution (CPD) can be compactly represented with a Conditional Probability Tree (CPTr), which takes advantage of Contextual Independencies. A Contextual Independence exists if a particular state of a parent node makes the other co-parents, i.e., the spouses, independent of the child node.

All unsupervised structural learning algorithms, including EQ, TabooEQ, and SopLEQ, are compatible with estimating probabilities with Conditional Probability Trees.

Usage

To use Conditional Probability Trees for estimation, select Edit > Parameter Estimation with Trees from the Menus. Alternatively, you can select Parameter Estimation with Trees from the Graph Contextual Menu.