Pseudolikelihood EM for Within-network Relational Learning

In this work, we study the problem of within-network relational learning and inference, where models are learned on a partially labeled relational dataset and then are applied to predict the classes of unlabeled instance in the same graph. Recent work in statistical relational learning has considered three alternative approaches for this setting: independent learning with independent inference, independent learning with collective inference, and collective learning with collective inference. Here independent refers to techniques that ignore the unlabeled data and collective refers to techniques that jointly consider the labeled and unlabeled data. Models from each of these categories has been employed previously in different settings, but to our knowledge there has been no systematic investigation comparing models from the three categories. In this paper, we develop a novel pseudolikelihood EM method that facilitates collective learning and collective inference on partially labeled n...

• Type: Conference
• Year: 2008
• Where: ICDM
• Authors: Rongjing Xiang, Jennifer Neville
• Keywords: Collective Inference, Data Mining, ICDM 2008, Independent Learning, Relational Learning