Username Remember Me?
Password   forgot password?
   
   
svm problem
Posted: 02 February 2009 08:40 AM   [ Ignore ]  
Newbie
Rank
Total Posts:  9
Joined  2009-02-02

I am new for prtools. I’m trying to use svmr, but it shows the following

?? Invalid MEX-file ‘C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll’:
C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll is not a
valid Win32 application.

I’m using windows xp

matlab 7.5.0 (R2007b)
prtools v 4.1.4

How do I fix this problem ?

Thanks

[ Edited: 06 February 2009 07:58 AM by pavel]
Profile
 
 
Posted: 02 February 2009 09:11 AM   [ Ignore ]   [ # 1 ]  
Administrator
Rank
Total Posts:  13
Joined  2008-05-02

Dear Dia,

welcome to our forum! I’m trying to recreate your problem. Can you run the PRTools regression example (prex_regr) or does it fail as well?
Could you send a code example that fails with this error?

Thanks,

Pavel

Profile
 
 
Posted: 03 February 2009 10:35 AM   [ Ignore ]   [ # 2 ]  
Newbie
Rank
Total Posts:  9
Joined  2009-02-02

Dear Pavel,

It also failed when I ran the the regression example (prex_regr).
It seems that the algorithm stop when optimization has to be used.
After I ran, it showed the following:

??? Invalid MEX-file ‘C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll’:
C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll is not a
valid Win32 application.

.

Error in ==> nu_svro at 112
v = qld (D, f, -A, b, lb, ub, p, length(b));

Error in ==> nu_svr at 122
[v,J,epsilon_or_nu] =
nu_svro(+K,y,C,svr_type,nu_or_epsilon,pd);

Error in ==> map at 115
d = feval(mapp,a,pars{:});

Error in ==> C:\Program
Files\MATLAB\R2007b\toolbox\prtools\@mapping\mtimes.p>mtimes
at 15

Error in ==> prex_regr at 15
w2a = a*nu_svr([],’r’,2,0.01,’epsilon’,0.1);

=============================================
Here is my svmr code.
I just tried to create a model.

trn = oc_set(trn);
model = svmr(trn,’r’);

after I ran, it showed the following

??? Invalid MEX-file ‘C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll’:
C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll is not a
valid Win32 application.

.

Error in ==> svmr at 65
alf = qld(H,f,Aeq,beq,lb,ub,[],1);

Error in ==> start at 6
model = svmr(trn,’r’);

Thanks

Profile
 
 
Posted: 03 February 2009 11:38 AM   [ Ignore ]   [ # 3 ]  
Administrator
Avatar
RankRankRankRank
Total Posts:  240
Joined  2008-04-26

Dear Dia,

I think there is several problems here combined. First, the ‘Invalid MEX’ error points to qld.dll which is very strange. Matlab 7.4 and higher should use the qld.mexw32 not .dll! Could you, please, check if your prtools/private directory contains the file qld.mexw32?

If not, please unzip there the one attached to this message and try simple SVM classifier training:

>> a=gendatb
Banana Set
100 by 2 dataset with 2 classes[50  50]
>> w=svc(a)
Support Vector Classifier2 to 2 trained classifier --> svc

Does the error message change now?

Thanks,

Pavel

[ Edited: 03 February 2009 11:40 AM by pavel]
File Attachments
qld.mexw32.zip  (File Size: 46KB - Downloads: 181)
Profile
 
 
Posted: 04 February 2009 10:31 AM   [ Ignore ]   [ # 4 ]  
Newbie
Rank
Total Posts:  9
Joined  2009-02-02

Dear Pavel,

I already have qld.mexw32 in prtools\private directory. Then, I ran the simple svm, but it still showed the following error:

??? Invalid MEX-file ‘C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll’:
C:\Program
Files\MATLAB\R2007b\toolbox\prtools\private\qld.dll is not a
valid Win32 application.

.

Error in ==> svo at 94
v = qld (D, f, -A, b, lb, ub, p, 1);

Error in ==> svc at 121
[v,J,C,nu] = svo(+K,y,C,Options);

Error in ==> start at 8
w=svc(a);

Thanks

Profile
 
 
Posted: 04 February 2009 11:35 PM   [ Ignore ]   [ # 5 ]  
Newbie
Rank
Total Posts:  9
Joined  2009-02-02

Dear Pavel,

Please ignore the previous message.

I think it works from now.

Thank you for your help and quick response.

Thanks

Dia,

Profile
 
 
Posted: 05 February 2009 07:56 AM   [ Ignore ]   [ # 6 ]  
Administrator
Avatar
RankRankRankRank
Total Posts:  240
Joined  2008-04-26

Dear Dia,

that’s great! Could you please share with us what was the issue? Was it the MEX?

Best Regards,

Pavel

Profile
 
 
Posted: 06 February 2009 01:42 AM   [ Ignore ]   [ # 7 ]  
Newbie
Rank
Total Posts:  9
Joined  2009-02-02

I think it was MEX file. I replaced the one that you attached to prtools/private. Then, it works
=============
I have a question about using svmr

my data set has 500 observations 8 features.

How do I define label in svmr since y is continuous values?

Thanks

Dia

Profile
 
 
Posted: 06 February 2009 07:58 AM   [ Ignore ]   [ # 8 ]  
Administrator
Avatar
RankRankRankRank
Total Posts:  240
Joined  2008-04-26

Dear Dia,

you need to define ‘targets’ - numerical values you will try to learn from you input data (targets are the y’s or dependent variables).
Maybe checkout example such as gendatsinc.m to see how to set up a dataset for regression.

Hope it helps,

Pavel

P.S.: I’m closing this thread. Please post different questions separately so that the content can be easily found. And keep asking :-) Thanks!

Profile