Uncertainty sampling and transductive experimental design for active dual supervision Export

@inproceedings{citeulike:5968883, abstract = {Dual supervision refers to the general setting of learning from both labeled examples as well as labeled features. Labeled features are naturally available in tasks such as text classification where it is frequently possible to provide domain knowledge in the form of words that associate strongly with a class. In this paper, we consider the novel problem of active dual supervision, or, how to optimally query an example and feature labeling oracle to simultaneously collect two different forms of supervision, with the objective of building the best classifier in the most cost effective manner. We apply classical uncertainty and experimental design based active learning schemes to graph/kernel-based dual supervision models. Empirical studies confirm the potential of these schemes to significantly reduce the cost of acquiring labeled data for training high-quality models.}, address = {New York, NY, USA}, author = {Sindhwani, Vikas and Melville, Prem and Lawrence, Richard D.}, booktitle = {ICML '09: Proceedings of the 26th Annual International Conference on Machine Learning}, citeulike-article-id = {5968883}, citeulike-linkout-0 = {http://portal.acm.org/citation.cfm?id=1553374.1553496}, citeulike-linkout-1 = {http://dx.doi.org/10.1145/1553374.1553496}, doi = {10.1145/1553374.1553496}, isbn = {978-1-60558-516-1}, keywords = {2008, active\_learning, experimental\_design, icml}, location = {Montreal, Quebec, Canada}, pages = {953--960}, posted-at = {2009-10-19 03:15:26}, priority = {2}, publisher = {ACM}, title = {Uncertainty sampling and transductive experimental design for active dual supervision}, url = {http://dx.doi.org/10.1145/1553374.1553496}, year = {2009} }

TY - CONF ID - citeulike:5968883 L3 - citeulike-article-id:5968883 N2 - Dual supervision refers to the general setting of learning from both labeled examples as well as labeled features. Labeled features are naturally available in tasks such as text classification where it is frequently possible to provide domain knowledge in the form of words that associate strongly with a class. In this paper, we consider the novel problem of active dual supervision, or, how to optimally query an example and feature labeling oracle to simultaneously collect two different forms of supervision, with the objective of building the best classifier in the most cost effective manner. We apply classical uncertainty and experimental design based active learning schemes to graph/kernel-based dual supervision models. Empirical studies confirm the potential of these schemes to significantly reduce the cost of acquiring labeled data for training high-quality models. CY - Montreal, Quebec, Canada SN - 978-1-60558-516-1 AD - New York, NY, USA EP - 960 TI - Uncertainty sampling and transductive experimental design for active dual supervision PB - ACM T2 - ICML '09: Proceedings of the 26th Annual International Conference on Machine Learning SP - 953 KW - 2008 KW - active_learning KW - experimental_design KW - icml AU - Sindhwani, Vikas AU - Melville, Prem AU - Lawrence, Richard PY - 2009/// UR - http://dx.doi.org/10.1145/1553374.1553496 ER -