Robust collective classification with contextual dependency network models

摘要: In order to exploit the dependencies in relational data improve predictions, classification models often need make simultaneous statistical judgments about class labels for a set of related objects. Robustness has always been an important concern such collective since many real-world as Web pages are accompanied with much noisy information. this paper, we propose contextual dependency network (CDN) model classifying linked objects presence and irrelevant links. The CDN makes use function characterize among so that it can effectively reduce effect links on classification. We show how Gibbs inference framework over multiple experiments demonstrates relatively high robustness datasets containing