Minimal Augmented Markov Blanket

The selection of variables realized with the Markov Blanket learning algorithm is based on a heuristic search. The set of selected nodes can be non-minimal, especially when there are various influence paths between the nodes and the target. In that case, the target analysis result takes into account too many nodes. By applying an unsupervised learning algorithm to the selected nodes, Minimal Augmented Markov Blanket learning reduces this set of nodes and results in a more accurate target analysis.

However, if the task is a pure prediction task (for example, a scoring function), the Augmented Markov Blanket algorithm is usually more accurate than its minimal version since it uses more “pieces of evidence.”