2002 ExploitingRelationsAmongConcept

From GM-RKB
Jump to navigation Jump to search

Subject Headings: Weakly Labeled Training Data.

Notes

Cited By

Quotes

Author Keywords

Abstract

We consider a learning setting in which there are well defined relations that exist among instances of certain classes. In particular, we consider the domain of predicting various types of gene-regulation elements in bacterial genomes. Given instances of one class, we can often acquire "weakly labeled" training data from another class by taking advantage of known relationships that exist between two classes. The examples are weakly labeled in that either the class label is incompletely specified, the exact extent of an instance is only partially specified, or both. We use an EM-based approach to handle this hidden state during learning. Our experimental results show that, when only small training sets are available, there can be significant value in augmenting the training sets with weakly labeled examples acquired from relationships among concepts.

References

,

 AuthorvolumeDate ValuetitletypejournaltitleUrldoinoteyear
2002 ExploitingRelationsAmongConceptMark Craven
Joseph Bockhorst
Exploiting Relations Among Concepts to Acquire Weakly Labeled Training Data