FLIP: active learning for relational network classification

摘要: Active learning in relational networks has gained popularity recent years, especially for scenarios when the costs of obtaining training samples are very high. We investigate problem active both single- and multi-labeled network classification absence node features during training. The becomes harder number labeled nodes available a model is limited due to budget constraints. inability use traditional setup data, motivated researchers propose Collective Classification algorithms that jointly classifies all test by exploiting underlying correlation between labels its neighbors. In this paper, we based on different query strategies using collective where each can belong either one class (single-labeled network) or multiple classes (multi-labeled network). have evaluated our method single-labeled networks, results promising cases several real world datasets.