Performance tuning options for rxDForest/rxDTree
|
Goal |
Options |
|
Performance |
xVal = 0 # (rxDTree) number of cross-validations during fittingcomputeOobError = -1 # (rxDForest) do not compute OOB prediction errorsscheduleOnce = TRUE # launch via rxExec (if modest data; best for Hadoop) # Requires nTree = #trees / #tasksblocksPerRead = large # if the input is an XDF file – Warning: reduce this if memory consumption is an issueverbose = 0 # minimize console outputreportProgress = 0 # dittouseSparseCube = FALSE # sparse cube is a bit slower so avoid if possible |
