Training an SVM from scratch!

Implements the sequential minimal optimization (SMO) algorithm introduced by J. Platt to solve the SVM dual problem. Currently configured to use either a linear or a gaussian kernel.