Few Advanced Parameters are set in the background file while few can be included as command line arguments.
-lc: print learning curve wth varying number of trees.
Parameters in Background file
Include the following lines in the background file (folder_bk.txt) to set the parameters:
Set the maximum number of nodes from root to leaf (height) in the tree.
Set the maximum number of literals in the node. Default value is 2.
Set the maximum number of clauses in the tree (i.e. maximum number of leaves). Default value is 100.
Set the maximum number of times the code should loop to learn clauses. Similar to numOfClauses but the counter increases even when no clause is learned. Default value is 100.
Allow reusing the literal from head of the clause in the body of the clause. Default is false.
Performs line search for deciding step length for functional gradient instead of using the fixed step length provided as
-step in basic parameters. Default value is false.
Prevent loading of all the existing libraries:
listsInLogic by setting it to false. Individual libraries can then be loaded using
importLibrary parameters. Default value is true.
setParam: loadAllLibraries = false.
Prevent loading of all the basic modes:
modes_listsInLogic by setting it to false. This might use a lot of cycles, so use with care. Default value is true
setParam: loadAllBasicModes = false.
Set the minimum number of trees used for printing learning curves. Used only when
-lc is set. Default value is 20
Set the number of trees to be increased every step while printing learning curve. Used only when
-lc is set. Default value is 2
Deprecated. Use maxTreeDepth instead
RDN Boost supports warm start, which allows you to add more trees to an already fitted model. To warm start learning, rename the existing
<target_predicate>.model file in the model directory to
<target_predicate>.model.ckpt and use the learn command as before.