The following code runs the algorithm from the paper: *Hyperplane folding - a way to increase the margin in Support Vector Machines Lars Lundberg, Håkan Lennerstad, Eva Garcia-Martin, Niklas Lavesson and Veselka Boeva
It does the following:
- Apply SVM in a dataset
- Select the primary support vector
- Divide the points into two, ones to the left of the support vector (forming an orthogonal line with the hyperplane) and the other ones to the right of the support vector
- Apply two SVM on each new set of points
- Rotate one side an angle a.
- The margin has thus increased from the initial one.