Gauge groups and data classification

摘要

We present a new method for constructing nonlinear classifiers. Given any two distinct sets of points in Rn the new method can construct, using gauge group techniques, a closed form expression of a surface, Φ(x)=0, which separates the two sets. We also show that any two distinct sets of points in Rn can be separated by a polynomial surface and present an algorithm for constructing such polynomial surfaces. Finally we present two numerical examples to illustrate the new method.