User’s Guide for the LinkConnectionStrength Package (Version 1.0) – A PNL Package for the Calculation and Vizualization of Entropy, Link Strengths and Connection Strengths in Discrete Bayesian Networks

Note: This may or may not be the most recent version of this document. The newest version is always available at www.DataOnStage.com. The LinkConnectionStrength package provides functions to calculate and vizualize entropy, connection strengths and link strengths for discrete Bayesian Networks. The package is imple-mented for Intel’s Open Source Probabilistic Network Library (PNL). The vizualization component relies on the GraphViz package to provide the actual picture of the graph. Within the graph varying gray scales of the links indicate link strengths and varying shades of the nodes indicate connection strengths relative to a specific node, while the actual numbers are provided as the labels of the links or nodes. The following measures are implemented: • Entropy is used to measure the uncertainty in a single node. • Mutual information is used to measure connection strength. • Two measures derived from mutual information are available to measure link strength (True Average Link Strength and Blind Average Link Strength). • In addition, mutual information percentage and link strength percentage are provided to measure the percentage of the existing uncertainty that has been removed. Complete definitions of all of the above terms are given in this document.