0

Training: [w, b] = vl_svmtrain(feature_matrix.FeatureVector, label_vector.labelsMatrix, 0.1) ;

Classification:
[~,~,~, scores] = vl_svmtrain(feature_matrix.FeatureVector_temp, labels, 0, 'model', w.w, 'bias', b.b, 'solver', 'none') ;

I have given the labels -1 and 1 while training. When I use the model for classification, it returns scores which are less than -1 and also greater than 1. How do I use these scores for binary classification? What does the score actually mean?

sheetal_158
  • 7,391
  • 6
  • 27
  • 44

1 Answers1

0

The scores should be positive for class '+1' and negative for class '-1'.

Basically, the scores correspond to the position of the classified data points in relation to the separating hyperplane learned by the SVM.

For more information on Support Vector Machines, look at the SVM fundamentals section in the VLFeat documentation, or many other online sources. For example these lecture slides are very illustrative.

Milan
  • 492
  • 4
  • 12