Exploration of Kernel Methods

We will observe how the choice of a kernel affects the capabilities of SVM.

• We will use Paint Data to paint some data.
• Connect its output to SVM
• Add the Predictions widget; its data should come from Paint Data, its model from SVM.
• Connect a Scatter Plot to Predictions; color by classes, shapes by SVM's predictions.

Set the SVM to use a polynomial kernel. We will change the degree of the polynomial (setting d).

1. Paint some data that cannot be separated with linear kernel (d=1), but is separable using the quadratic kernel (d=2)
2. Paint some data the requires d=3.
3. What about this? Also, try it with RBF kernel.