These instances, whenmapped to an N-dimensional space, represent a core set that can be used to construct an approximation to theminimumenclosing ball. Solving the SVMlearning problem on these core sets can produce a good approximation solution i