pvnick / rf-ace

Automatically exported from code.google.com/p/rf-ace
0 stars 0 forks source link

Large forests fail to be built #35

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago
What steps will reproduce the problem?
1. Use a large trainset (48 features, 270000 instances)
2. Build forest with ntrees=300, mtry=12 features
3. Fails to be built, stacktrace is attached, but will not always be dumped

What version of the product are you using? On what operating system?
version r513
rf-ace-build-predictor.exe (built with gcc and make)
rf-ace-build-predictor-win64.exe (built with Visual Studio Express 10)

Please provide any additional information below.
I tried also 100 and 200 trees, which worked just fine, but with 300 there 
seems to be an issue. I don't think it's RAM related, since there were about 
6Gig free RAM left.

Original issue reported on code.google.com by berni.le...@gmail.com on 6 Apr 2012 at 7:10

Attachments:

GoogleCodeExporter commented 8 years ago
Thanks for providing the information. I will look into this once the CART 
inaccuracy and OOB error behavior as function of mTry are investigated.

Original comment by timo.erk...@gmail.com on 8 Apr 2012 at 4:22

GoogleCodeExporter commented 8 years ago
This should be resolved by now, at least I've successfully grown similar-sized 
forests in 32-bit environment with less than 2GB of RAM. Closing...

Original comment by timo.erk...@gmail.com on 11 Mar 2013 at 4:38