Key points are not available for this paper at this time.
We propose a family of learning algorithms based on a new form of regularization that allows us to exploit the geometry of the marginal distribution. We focus on a semi-supervised framework that in...
BelkinMikhail et al. (Fri,) studied this question.