The aim of this app is to use a traditional Support Vector Machine for modeling the surface of an object. The surface of the object is defined by a set of points and links, as shown in the left figure of the teapot (see upper figures). On the left a nu-SVM classifier was used to model the teapot of the University of Utah. The model shown in gray is the classifier’s decision surface. Note how the method preserves the object topology. On the right of the same figure the object shown is also a component of the teapot of the University of Utah. A epsilon-SVR regression machine is used to model a discrete function in the three dimensional space, the zero level surface of this function is used to model this object.
The rabbit of the University of Stanford is also showed in the upper figures, the left figure has been colored with the position error. At the right a model has been obtained using 3,219 support vectors for zero error of classification.