CvBoost train with updating (Bug #3208)

Added by Maria Dimashova over 11 years ago. Updated over 9 years ago.

Status:	Open	Start date:	2013-08-10
Priority:	Normal	Due date:
Assignee:	Vadim Pisarevsky	% Done:	0%
Category:	ml
Target version:	3.0
Affected version:	branch '2.4'	Operating System:	Any
Difficulty:		HW Platform:	Any
Pull request:

Description

After I fixed the segfault from #3193 on my PC, I faced that CvBoost doesn't do updating in the train method actually because of incorrect data_root->num_valid array (does not change the forest at all)... One more fix. After that the updating started to run. But the algorithm of updating looks strange for me. It trains new trees independently of previous ones; in prediction method it just sums values of trees from the both parts of the forest. I think this is wrong.
We should implement correct updating of CvBoost when it'll be refactored (now it's hard even to support the code).