Transductive Rademacher Complexity and Its Applications

We present data-dependent error bounds for transductive learning based on transductive Rademacher complexity. For specific algorithms we provide bounds on their Rademacher complexity based on their “unlabeled-labeled” decomposition. This decomposition technique applies to many current and practical graph-based algorithms. Finally, we present a new PAC-Bayesian bound for mixtures of transductive algorithms based on our Rademacher bounds.

• Type: Conference
• Year: 2007
• Where: COLT
• Authors: Ran El-Yaniv, Dmitry Pechyony
• Keywords: COLT 2007, Data-dependent Error Bounds, Machine Learning, Transductive Learning, Transductive Rademacher Complexity