In the literature, for a binary classification problem, I have come across examples where a One-ClassSVM model is trained using the data for only one of the two training labels and sometimes using both the training labels.
Which is the correct way to do this?
Train the model using both the classes or use only one of the class and test using the other class?
I have found these links to be related but don't answer my question:
(1) Best way to train one-class SVM
(2) what would be a recommended division of train and test data for one class SVM?
