Matlab died when randomforest finished its process about randomforest-matlab HOT 15 OPEN

can you give the part of the data and code for which this happens?

also did you run randomforest before in the same matlab instance and cancel it 
by any chance 
via ctrl+C?

thanks

[deleted comment]

I ran below codes, but it happens again.

>> train_data = rand(10,100);
>> train_label = [ones(5,1); ones(5,1)*-1];
>> Model = classRF_train(train_data, train_label, 0, 0);

Yes, I think I canceled job with ctrl+C.
However, I logged out and ran it again, but still it happens.

Thanks. 

well i am not able to reproduce the problem on linux with ssh(ing) from the 
same machine. And 
also from windows to a linux machine via putty

i can suggest the following: try using putty as a ssh client or use `screen' 
command to spawn 
matlab and see.

btw is the X being forwarded on windows via SSH? try using without the 
forwarding

I just found that this happens when I run matlab with nodisplay option.
In the command line.
>> matlab -nodisplay

I run the matlab on console mode then it happens. 

Thanks.

sorry i could not get to you earlier.

if it is same to you, i.e getting matlab on ssh, try with matlab -nojvm and 
with a different ssh 
client

i am still not able to reproduce the same error on my machine with and without 
ssh and using 
the nodisplay option

I tried -nojvm with putty but the problems is still the same.

Thanks.

I too am getting exactly the same error as korjcjeong. I am using 7.5.0.342 
(R2007b).
In particular, the problem only happens for classification (i.e. when using 
classRF_train.m) and not for regression (i.e. refRF_train.m is OK).  I wonder 
if there is an issue using calloc [e.g. double* countts = (double*) 
calloc(nclass * nts,sizeof(double)); on line 189] instead of letting MATLAB 
doing the memory managment, as is recommended, or a memory leak.

so are you also using it in the same setting of ssh's and using matlab? i was 
never able to reproduce the same problem.


well matlab variables cannot be used in the C code if they are to be written 
to, they are good only for reading, thats why there are allocs in C.

Hi,

I also have the same problem. Matlab freezes after it returns from 
classRF_train when I run matlab through ssh without GUI.
I tried this on SUSE and REDHAT and both get stuck after classRF_train and also 
classRF_predict.
It works without problem when I run matlab with GUI.

Thanks,
MJ

@MJ

did you try out my earlier suggestion in the other thread?

possible issue might be that particular version of gfortran and g++ don't work 
well in conjunction with malloc and matlab. try to sync with the svn source 
which replaces mallocs by matlab based malloc mxMalloc 
http://code.google.com/p/randomforest-matlab/source/checkout

I have solved this problem by disabling X11 tunneling.

[deleted comment]

I forgot to mention that, yes I updated gfortran and g++ to 4.4.1 version and I 
use the latest version on the repository but still I have the problem. I'm 
trying to reproduce the problem to find out more about the issue. I'll provide 
more information if I find any.

Thanks,
MJ

hi mj

what options do you use to run matlab in ssh?  korjc (thanks!) reported it 
might be due to X11 tunneling. the other option is to see if one or a 
combination of the startup options is causing issue (comment-5), startup 
options include either -nojvm -nodisplay

if you a test code i can try it on my machines and try to see if i can 
reproduce them on debian.