Gaussian Bayes Network Structure Learning tool
The application has a console input, where many of the parameters can be specified. Here are a few simple examples:
java –jar GBNStructureLearner.jar –df res/sample.0.data.csv –sf res/sample.0.structure –sa hillclimbing –re 100 –dr 10000
java –jar GBNStructureLearner.jar –df res/sample.0.data.csv –sf res/sample.0.structure –sa simulatedannealing –ns 100000 –lb 0.2
java –jar GBNStructureLearner.jar -d res/sample.0.data.csv -st res/sample.0.structure -e -ld myNetwork.txt
The complete list of options are the following:
Option | Long option | Parameter? | Desciption |
---|---|---|---|
-df | data-filename | Yes | specify file containing data |
-dr | data-rows | Yes | specify the first how many rows should be used for searching |
-eo | evaluation-only | No | if you only wish to evaluate the network |
-lb | lambda | Yes | specify how strong the regularization should be (default=0) |
-ld | load-from-file | Yes | if you wish to load the network you saved previously, specify the file |
-ns | number-of-steps | Yes | specify the number of steps the search algorithm should make (default=10000) |
-re | random-edges | Yes | specify how many random edges should the graph contain (default=0) |
-sa | search-algorithm | Yes | choose searching algorithm (default=sa) (options: sa/simulatedannealing/hc/hillclimbing) |
-sf | structure-filename | Yes | specify file containing network structure |