0
$\begingroup$

I'm training a classification SVM for a binary classification model with 20 features (LIBSVM). These are some reports of my model (with different combination of features). Why all nBSVs are zero? Otherwise, based on *.* and .* outputs, the SVM trains real fast. Is this a problem?

.*
optimization finished, #iter = 1194
nu = 0.007242
obj = -296.242686, rho = -0.026875
nSV = 617, nBSV = 0
Total nSV = 617
*.*
optimization finished, #iter = 924
nu = 0.007236
obj = -296.473015, rho = 0.002460
nSV = 618, nBSV = 0
Total nSV = 618
.*
optimization finished, #iter = 1204
nu = 0.007288
obj = -298.117823, rho = -0.021878
nSV = 618, nBSV = 0
Total nSV = 618
.*
optimization finished, #iter = 1169
nu = 0.007248
obj = -296.465051, rho = -0.012346
nSV = 618, nBSV = 0
Total nSV = 618
.*
optimization finished, #iter = 1449
nu = 0.007177
obj = -367.201873, rho = -0.015914
nSV = 772, nBSV = 0
Total nSV = 772
.*.*
optimization finished, #iter = 1355
nu = 0.008642
obj = -300.306592, rho = -0.028429
nSV = 617, nBSV = 0
Total nSV = 617
.*
optimization finished, #iter = 1196
nu = 0.008608
obj = -298.631753, rho = -0.017603
nSV = 616, nBSV = 0
Total nSV = 616
*.*
optimization finished, #iter = 774
nu = 0.008684
obj = -301.772061, rho = 0.000564
nSV = 618, nBSV = 0
Total nSV = 618
.*.*
optimization finished, #iter = 1313
nu = 0.008646
obj = -299.953007, rho = -0.026729
nSV = 617, nBSV = 0
Total nSV = 617
.*
optimization finished, #iter = 1107
nu = 0.008649
obj = -300.051120, rho = -0.007149
nSV = 617, nBSV = 0
Total nSV = 617
.*
optimization finished, #iter = 1482
nu = 0.008603
obj = -373.305016, rho = -0.013622
nSV = 771, nBSV = 0
Total nSV = 771

From LIBSVM documentation:

obj is the optimal objective value of the dual SVM problem. rho is the bias term in the decision function sgn(w^Tx - rho). nSV and nBSV are number of support vectors and bounded support vectors (i.e., alpha_i = C). nu-svm is a somewhat equivalent form of C-SVM where C is replaced by nu. nu simply shows the corresponding parameter.

$\endgroup$
3
$\begingroup$

It means that your training loss is 0 (i.e., your data is separable).

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.