Name
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. |
Year | DOI | Venue |
|---|---|---|
2006 | 10.1145/1143844.1143980 | ICML |
Keywords | Field | DocType |
extensive experimental result,preference learning,efficient optimization,measurements y,classic experimental design method,active learning,transductive experimental design,new method,questionnaire design,new design problem,design method,experimental design | Transduction (machine learning),Semi-supervised learning,Active learning,Active learning (machine learning),Pattern recognition,Computer science,Design methods,Greedy algorithm,Artificial intelligence,Preference learning,Machine learning,Scalability | Conference |
ISBN | Citations | PageRank |
1-59593-383-2 | 93 | 3.92 |
References | Authors | |
13 | 3 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Yu, Kai | 1 | 4799 | 255.21 |
| Jinbo Bi | 2 | 1432 | 104.24 |
| Volker Tresp | 3 | 2907 | 373.75 |