Relational Active Learning for Joint Collective Classification Models

In many network domains, labeled data may be costly to acquire - indicating a need for relational active learning methods. Recent paper has demonstrated that relational model performance can be improved by taking network structure into account when choosing instances to label. However, in collective inference settings, both model estimation and prediction can be improved by acquiring a node's label - since relational models estimate a joint distribution over labels in the network and collective classification methods propagate information from labeled training data during prediction.

Provided by: Purdue Federal Credit Union Topic: Security Date Added: May 2011 Format: PDF

Find By Topic