Noise-tolerant rule induction from multi-instance data
| Author(s) : | Jean-daniel Zucker Yann Chevaleyre, |
| Publisher : | N/A |
| Publication Date : | 2000 |
| ISSN : | N/A |
| Abstract : | This paper addresses the issue of multipleinstance induction of rules in the presence of noise. It first proposes a multiple-instance extensions of rule-based learning algorithms. Then, it shows what kind of noise can appear in multiple-instance data, and how to handle it theoretically. Finally, it describes the implementation of such a noise-tolerant multiple instance learner, and shows its performance on several problems, including the well-known mutagenesis prediction task. 1., |