Graph Based Classication Methods Using Inaccurate External Classier Information

In this paper we consider the problem of collectively classifying entities where relational information is available across the entities. In practice inaccurate class distribution for each entity is often available from another (external) classier. For example this distribution could come from a classier built using content features or a simple dictionary. Given the relational and inaccurate external classier information, we consider two graph based settings in which the problem of collective classication can be solved. In the rst setting the class distribution is used to x labels to a subset of nodes and the labels for the remaining nodes are obtained like in a transductive setting. In the other setting the class distributions of all nodes are used to dene the tting function part of a graph regularized objective function. We dene