Large Benchmark Dataset

#Large Benchmark Dataset

##Description

This dataset has been provided by bnlearn repository (an R package for Bayesian network learning and inference). This dataset is used for benchmarking the PGMLab application and to show the speed at which the tool can perform learning and inference on a typical large dataset.

###Files

Large benchmark dataset can be found in the folder "data/munin-dataset/"

####Pairwise interaction file

The pairwise interaction file can be use to create a logical factorgraph using either the R or C interface. It is used to let the PGMLab know which nodes to index when performing learning and inference as well.

####Observed Data files

Name	Percent of Nodes Observed
visibleSet_0.5.txt	50%
visibleSet_0.7.txt	70%
visibleSet_0.9.txt	90%

####Generating Learnt Factorgraph file - Learning (estimated parameters)

In order to generate a learnt factorgraph you will need to use the logical factorgraph that you created from the pairwise interaction file, in combination with one of the three observed data files.

####Generating Posterior probabilities file - Inference

When generating the posterior probabilities you can either use a logical or learnt factorgraph in combination with one of the three observation files.