Performance tuning options for rxDForest/rxDTree
| Goal | Options |
| Performance | xVal = 0 # (rxDTree) number of cross-validations during fitting computeOobError = -1 # (rxDForest) do not compute OOB prediction errors scheduleOnce = TRUE # launch via rxExec (if modest data; best for Hadoop) # Requires nTree = #trees / #tasks blocksPerRead = large # if the input is an XDF file – Warning: reduce this if memory consumption is an issue verbose = 0 # minimize console output reportProgress = 0 # ditto useSparseCube = FALSE # sparse cube is a bit slower so avoid if possible |