- [Home]
- [Research achievement]
- [Research achievement detail]
Title | The Classification Problem in Generalized Label Noise Model (in Japanese) |
---|---|
Authors | Tota Suko 、Shunsuke Horii |
Released Year | 2017 |
Format | Conference |
Category | Knowledge information processing |
Jounal Name | |
Jounal Page | vol. IEICE-117, no.293, pp.377-382 |
Published Year | 2017 |
Published Month | 11 |
Abstract (English) |
In classification problem, there is a case where noise is added to the label. In this study, we proposes a general noise added model. We describe a variety of learning problems using the proposed model. For example, semi-supervised learning, learning from positive and unlabeled data and learning from data including the outliers. For this model, we propose a classification algorithm using the EM algorithm and the Variational Bayes method. We evaluate the performance of the proposed algorithm by numerical experiments. |
Note (English) |
1 |
Manuscript | |
Presentation |