The quadratic programming routine in a support vector machine solver provides the basis for splitting the training examples into three categories. Define these three categories, and use a two-dimensional figure to illustrate how the splitting is performed.