I\'m trying to find training methods of a threshold setting based on one-class support vector machine. So my question is, is decision surface as hypersphere in one class svm