Figure 7.
Principle of the linear support vector machine, which lifts the indeterminacy of separable classification by fitting the thickest margin, stuck in between a few ‘support vectors’. The principle can be extended to nonlinear class separation by using Mercer kernels [125].