Active learning via transductive experimental design Export

@inproceedings{citeulike:5968885, abstract = {This paper considers the problem of selecting the most informative experiments x to get measurements y for learning a regression model y = f (x). We propose a novel and simple concept for active learning, transductive experimental design , that explores available unmeasured experiments (i.e., unlabeled data) and has a better scalability in comparison with classic experimental design methods. Our in-depth analysis shows that the new method tends to favor experiments that are on the one side hard-to-predict and on the other side representative for the rest of the experiments. Efficient optimization of the new design problem is achieved through alternating optimization and sequential greedy search. Extensive experimental results on synthetic problems and three real-world tasks, including questionnaire design for preference learning, active learning for text categorization, and spatial sensor placement, highlight the advantages of the proposed approaches.}, address = {New York, NY, USA}, author = {Yu, Kai and Bi, Jinbo and Tresp, Volker}, booktitle = {ICML '06: Proceedings of the 23rd international conference on Machine learning}, citeulike-article-id = {5968885}, citeulike-linkout-0 = {http://portal.acm.org/citation.cfm?id=1143980}, citeulike-linkout-1 = {http://dx.doi.org/10.1145/1143844.1143980}, doi = {10.1145/1143844.1143980}, isbn = {1-59593-383-2}, keywords = {2006, active\_learning, experimental\_design, icml}, location = {Pittsburgh, Pennsylvania}, pages = {1081--1088}, posted-at = {2009-10-19 03:17:22}, priority = {2}, publisher = {ACM}, title = {Active learning via transductive experimental design}, url = {http://dx.doi.org/10.1145/1143844.1143980}, year = {2006} }

TY - CONF ID - citeulike:5968885 L3 - citeulike-article-id:5968885 N2 - This paper considers the problem of selecting the most informative experiments x to get measurements y for learning a regression model y = f (x). We propose a novel and simple concept for active learning, transductive experimental design , that explores available unmeasured experiments (i.e., unlabeled data) and has a better scalability in comparison with classic experimental design methods. Our in-depth analysis shows that the new method tends to favor experiments that are on the one side hard-to-predict and on the other side representative for the rest of the experiments. Efficient optimization of the new design problem is achieved through alternating optimization and sequential greedy search. Extensive experimental results on synthetic problems and three real-world tasks, including questionnaire design for preference learning, active learning for text categorization, and spatial sensor placement, highlight the advantages of the proposed approaches. CY - Pittsburgh, Pennsylvania SN - 1-59593-383-2 AD - New York, NY, USA EP - 1088 TI - Active learning via transductive experimental design PB - ACM T2 - ICML '06: Proceedings of the 23rd international conference on Machine learning SP - 1081 KW - 2006 KW - active_learning KW - experimental_design KW - icml AU - Yu, Kai AU - Bi, Jinbo AU - Tresp, Volker PY - 2006/// UR - http://dx.doi.org/10.1145/1143844.1143980 ER -