In this paper. we propose TRAM. a transductive multi-label classification method by label set propagation. At first. we formulate the task as an optimization problem which is able to exploit unlabeled data to obtain an effective model for assigning appropriate multiple labels to instances. Then. we develop an efficient algorithm which has a closed-form solution for this optimization problem. Empirical studies on a broad range of real-world tasks demonstrate that our TRAM method can effectively boost the performance of multi-label classification by using unlabeled data in addition to labeled data.
Check Also
North America’s Microchip Revolution Is Underway
As the U.S. strives to develop stronger regional supply chains for several key components supporting …