How does SVM classify work?

I am absolutely new to SVM or any of these classification techniques. Now i was learning how to use SVM-multiclass to classify data, and I am confused:

I understand perfectly how the svm-learn works on the training data, by creating hyperplanes and stuff, and finding support-vectors.

What I dont seem to get is how svm-cla开发者_C百科ssify works, or rather, What is its actual function? From its name, svm-classify should "assign classes to the unclassified points" but it seems to just give me the "error" and "average-loss" on test set.

to be more clear :

if i train the svm with this file:

class-label : data

and then i pass the test file like this:

    data1
    data2
    .
    .
    dataN

so svm_classify should output the classes to these data...

Shouldn't it ?

Looks like you are using svm-light but not libsvm. After you train the data, you should have a model file. Then you run the svm_classify as follows:

svm_classify [options] example_file model_file output_file

The classification result should be written in output_file.