Abstract | ||
---|---|---|
Multi-view algorithms reduce the amount of re- quired training data by partitioning the domain features into separate subsets or views that are sufficient to learn the target concept. Such al- gorithms rely on the assumption that the views are sufficiently compatible for multi-view learn- ing (i.e., most examples are labeled identically in all views). In practice, it is unclear whether or not two views are sufficiently compatible for solving a new, unseen learning task. In order to cope with this problem, we introduce a view validation algorithm: given a learning task, the algorithm predicts whether or not the views are sufficiently compatible for solving that partic- ular task. We use information acquired while solving several exemplar learning tasks to train a classifier that discriminates between the tasks for which the views are sufficiently and insuffi- ciently compatible for multi-view learning. Our experiments on wrapper induction and text clas- sification show that view validation requires only a modest amount of training data to make high accuracy predictions. |
Year | Venue | Keywords |
---|---|---|
2002 | ICML | automatic view detection,adaptive view validation,first step |
Field | DocType | ISBN |
Training set,Pattern recognition,Computer science,Artificial intelligence,Classifier (linguistics),Machine learning,Exemplar learning | Conference | 1-55860-873-7 |
Citations | PageRank | References |
23 | 1.34 | 8 |
Authors | ||
3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Ion Muslea | 1 | 1344 | 121.66 |
Steven Minton | 2 | 3473 | 536.74 |
Craig A. Knoblock | 3 | 5229 | 680.57 |