This program is an implementation of the method presented in the paper
Log-linear models independence structure comparison,
see https://arxiv.org/abs/1907.08892
$ java -jar jllcm.jar model1 model2 n [method]
model1
and model2
are Markov network (.mn) files in the format for Libra Toolkit. This implementation supports only binary variables.
n
is the number of variables
[method]
is an optional argument: bf
for brute force; both
for default and brute force; omit for default method;
slowbf
uses the first implementation of brute force which generates all contexts and is more computationally expensive.
The output shows the values for the confusion matrix. Example output:
LLCM: [TP: 16, FN: 8; FP: 0, TN: 0]
BF: [TP: 16, FN: 8; FP: 0, TN: 0]
For examples and more information see examples/README and examples/README_featureDifference.