Problem with Random Forest Learner (Regression): flow variable for minimum child and node size (KNIME 4.3.2 linux)

[davidrialfigols]

Hi!
I’m relatively new in Knime. I am trying to optimize the parameter “minimum node size” of the Random Forest Learner (Regression) node using flow variables and a parameter optimization loop. When I configure the flow variables of the RFL an error appears with the message “errors loading flow variables into node: Invalid minimum child size (2); must be at most 2 x minimum node size (2)” which obviously 2 < 2x2, so at least the message error is incorrect.

Thanks in advance
David

[ScottF]

Hi @davidrialfigols and welcome to the forum. Sorry for letting this one fall off the radar a little bit.

I checked and am able to reproduce the warning message that you are seeing. I also noticed that strangely there are two flow variables in the Random Forest Learner - minNodeSize and minChildSize - but only one of those has a corresponding option in the configuration dialog. This is probably a holdover from the Tree Ensemble Learner.

At any rate, it looks like there are some issues we need to tidy up with this node. Thanks for your feedback!

EDIT: Ticket created (AP-16533).

[sudheerkopparapu]

Hello @ScottF

Is this issue now fixed in RF learner (Regression)? I am also having the same problem now.

Thanks

[ScottF]

Hi @sudheerkopparapu -

Sorry for the trouble. I checked the ticket, but there is not progress to report yet. We will sure to post in this thread if there is an update.