Support Vector Classifiers
- We can’t always use a hyperplane to separate two classes.
- Even if such a classifier does exist, it’s not always desirable, due to overfitting or too much sensitivity to individual observations.
- Thus, it might be worthwhile to consider a classifier/hyperplane that misclassifies a few observations in order to improve classification of the remaining data points.
- The support vector classifier, a.k.a the soft margin classifier, allows some training data to be on the wrong side of the margin or even the hyperplane.