nehalg1 / randomforest-matlab

Automatically exported from code.google.com/p/randomforest-matlab
0 stars 0 forks source link

number of trees #14

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago
I am using MATLAB 7.5 on Windows 7

I am try to use the code and use 2500 random forests for my training set. And 
it runs out of memory. 
So my question is: is the code retaining ALL the trees during the training or 
should it (or does it) only just retain the best one so far? The second option 
will not cause memory issues.

This is related to file classRF.cpp: and specifically the line: 

for(jb = 0; jb < Ntree; jb++) {

i will be grateful for the reply,

Original issue reported on code.google.com by imra...@gmail.com on 7 Mar 2011 at 11:41

GoogleCodeExporter commented 8 years ago
the forest retains all the trees. technically there is no best tree as all the 
trees are from the same distribution and thus will perform equally good.

the issue might lie with your OS. if its 32bit then it doesnot allow more than 
2 GB of space per process and in this case matlab

i do know that the forest matrices are huge and grow proportionally to the 
number of examples you have. But on my 64 bit machine with 12gb ram i can grow 
500 trees for data exceeding 200K examples with > 500 features

Original comment by abhirana on 7 Mar 2011 at 5:53

GoogleCodeExporter commented 8 years ago

Original comment by abhirana on 31 Mar 2012 at 8:39