Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

LatentSVM: svmlight training parameters

I am trying to train a model for svm using HOG features. I have created the features file from positives & negatives. The function SVMlight::getInstance()->read_problem("genfiles\features.dat") executes without errors (Following is the output):

{ Reading files, generating HOG features and save them to file 'genfiles/features. dat': Calling SVMlight Scanning examples...done Reading examples into memory...100..200..300..400..500..600..700..800..900..OK. (936 examples read) }

However, when I call: { SVMlight::getInstance()->train(); }

It fails with the following output: { Optimizing...........................................................done. (60 i terations) Optimization finished (maxdiff=0.00098). Runtime in cpu-seconds: 2.09 Number of SV: 75 (including 1 at upper bound) L1 loss: loss=0.02505 Norm of weight vector: |w|=0.32211 Norm of longest example vector: |x|=14.71081 Number of kernel evaluations: 16574 Writing alpha file... ══════════════════════════════════════════════════════════ ════════════════════════════════════════════════════════════════════════════════ ═════════════════════════════════════════════════════════════════════════▬Vτ₧»♥╥ <════════════════════════: No such file or directory }

I debugged it and found that the alpha file path in the function "write_alphas" is garbage. May be because I have not passed any parameters to the svmlight (LEARN_PARM* learn_parm is not set to any valid values?).

Could anybody please provide some directions on these parameters?

Thanks.

LatentSVM: svmlight training parameters

I am trying to train a model for svm using HOG features. I have created the features file from positives & negatives. The function SVMlight::getInstance()->read_problem("genfiles\features.dat") executes without errors (Following is the output):

{

Reading files, generating HOG features and save them to file 'genfiles/features.
dat':
Calling SVMlight
Scanning examples...done
Reading examples into memory...100..200..300..400..500..600..700..800..900..OK.
(936 examples read)
}

However, when I call: { SVMlight::getInstance()->train(); }

call:

SVMlight::getInstance()->train();

It fails with the following output: { output:

Optimizing...........................................................done. (60 i
terations)
Optimization finished (maxdiff=0.00098).
Runtime in cpu-seconds: 2.09
Number of SV: 75 (including 1 at upper bound)
L1 loss: loss=0.02505
Norm of weight vector: |w|=0.32211
Norm of longest example vector: |x|=14.71081
Number of kernel evaluations: 16574
Writing alpha file... ══════════════════════════════════════════════════════════
════════════════════════════════════════════════════════════════════════════════
═════════════════════════════════════════════════════════════════════════▬Vτ₧»♥╥
<════════════════════════: No such file or directory
}

I debugged it and found that the alpha file path in the function "write_alphas" is garbage. May be because I have not passed any parameters to the svmlight (LEARN_PARM* learn_parm is not set to any valid values?).

Could anybody please provide some directions on these parameters?

Thanks.