Learning with Local and Global Consistency

We consider the general problem of learning from labeled and unlabeled data, which is often called semi-supervised learning or transductive inference. A principled approach to semi-supervised learning is to design a classifying function which is sufficiently smooth with respect to the intrinsic structure collectively revealed by known labeled and unlabeled points. We present a simple algorithm to obtain such a smooth solution. Our method yields encouraging experimental results on a number of classification problems and demonstrates effective use of unlabeled data.

• Type: Conference
• Year: 2003
• Where: NIPS
• Authors: Dengyong Zhou, Olivier Bousquet, Thomas Navin Lal, Jason Weston, Bernhard Schölkopf
• Keywords: NIPS 2003, NIPS 2007, Semi-supervised Learning, Transductive Inference, Unlabeled Data