Random forest in TMVA

In the tmva user manual (TMVA version 4.3.0 for ROOT >= 6.12/00, dated May 26, 2020), section 8.13.3, it is stated that Randomised Trees (like the Random Forests) is available through the option UseRandomizedTree.
However, based on my understanding, random forest is not used with adaboost (and other boosting methods). my questions are:

  1. how could i do the “true” random forest, when the BoostType available are only AdaBoost, RealAdaBoost,AdaBoostR2,Grad ?
  2. or i just misunderstood the manual, and random forest is not implemented in tmva, and what the training option UseRandomizedTree is only to randomly choose the variables at each step of the boosting (and not building the true random forest) ?

additional question:
the user manual i have is for TMVA version 4.3.0 (for ROOT >= 6.12/00 on May 26, 2020)
but the manual seems to be a little bit different from my current version (for example, the options available for a particular machine learning model).

is there any updated user manual, or portals (i went to the portals like sourforce, and even root/tmva on cern and they just not updated or dead!) that provide guides on the options available for a particular machine learning model.

_ROOT Version: 6.24 (PyROOT via conda)
_Compiler: gcc9

I think @moneta can help here


I am not sure about the answer to your first question, probably you are correct in saying that standard random forest is not available. I would need to ask to the author of the decision tree in TMVA.

For the user manual, version of May 2020 is the latest one. If you see some options are not documented there please let us know and we will fix the manual.
The source forge portal has not been updated since a lot of time.



yes, I am referring to the standard random forest, which involve both randomizing the data and the set of variables used to make cut (when building a tree).

For the user guidd, I just wondering that if the BDT (in ROOT 6.24) still use the same set of available parameters as in the latest manual. The available options are not well documented, and I have to rely on the user guide of the previous version.

Sometimes it leads to confusion. For example, I am not sure if the example in the Classification.C has listed all the available options for the BDT.
The other confusion I have encountered is that there is an option UseBaggedBoost in the BDT, but this option (and its explaination) is not listed in the current version of user guide. So I am not sure what that option does.

Thank you for the feedback, I will check and update the UsersGuide if needed