JMLR
2006
14 years 6 months ago
2006
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...
ICDM
2005
IEEE
14 years 11 months ago
2005
IEEE
In this paper, we propose a semi-supervised framework for learning a weighted Euclidean subspace, where the best clustering can be achieved. Our approach capitalizes on user-const...