Training Files for Structure Learning

The DataGenerator feature in the BBN module can be used to simulate the network and generate a data file from the network. The output data file can be used as the training dataset of any learning algorithm. The simulation process is done by forward sampling, i.e., first simulating root nodes, then simulating their children, and so on. It is included in the bbn package. DataGenerator2.java is an optimized version of DataGenerator.java. The output data file is stored in an xml file format. DataGenerator is also used to generate the evidence file if the number of samples is set to zero. The following example shows how to simulate the network and generate a data file with 20 samples, named asia20.xml, from the network. You can view the resulting data file here.

============================= Example of using DataGenerator ====================================

D:\bnj>java bbn/DataGenerator2
 * * * DataGenerator2 * * *
Usage: java DataGenerator2 input.xml output.xml number_of_samples (0 will 
generate an evidence file)

D:\bnj>java bbn/DataGenerator2 asia.xml asia20.xml 20
 * * * DataGenerator * * *
Usage: java DataGenerator input.xml output.xml number_of_samples (0 will 
generate an evidence file)
Success!

D:\bnj>

============================= End DataGenerator Example =========================================

KDD-Tools
Fri June 28 2002