Closed adamjohnwright closed 8 years ago
Made it link to split_nodes.pl file. This runs every time a job is submitted from PGMLab website. The old and new pi file are kept in the package that is provided in the results package. And information on the split is in the pgmlab.log file.
The problem is that the algorithm uses an exponential amount of memory when storing the CPT for each node. This increases with the number of parents. We would like to devise a way to have PGMLab handle these circumstances automatically (without the user needing to be aware).