Question

I have been using the lib-svm binaries for classification of my data.

For training I use the below command

`svm-train.exe -s 0 -t 2 -c 0.03125 -g 0.25 file.train file.train.model’

and get the following results

optimization finished, #iter = 10
nu = 1.000000
obj = -0.625000, rho = -0.000000
nSV = 20, nBSV = 20
Total nSV = 20

Now since I am using the rbf kernel in training set do I not have to include that in test data ?

`svm-predict.exe -s 0 -t 2 -c 0.03125 -g 0.25 file.test file.train.model file.out’

and i get

Unknown option: -s

Using the above command without any parameters works fine….

But I wanted to know that does libsvm automatically use the rbf kernel for testing ?

Answer 1

libsvm stores all the necessary information in the model file that is created from training, so no you don’t need to include those parameters, it will know what kernel and parameters to use. You only need to supply the model file and the test data.