| Presentation | 2019-07-26 Unbiased Estimation Equation for f-Separable Bregman Distortion Measures and the Properties of Its Estimators Masahiro Kobayashi, Kazuho Watanabe, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In this study, we discuss unbiased estimation equations in a class of objective function using the monotonically increasing function $f$ and Bregman divergence. The choice of the function $f$ gives desirable properties, such as robustness against outliers.However generally it does not guarantee the consistency of the estimators because it does not satisfy unbiased estimation equations. In order to satisfy unbiased estimation equations, analytically intractable integrals are generally required as bias correction terms. In this study, we clarify the combination of Bregman divergence, probability model, and function $f$ in which the bias correction term vanishes. Specifically, we investigate the case of squared and Itakura-Saito distances. In the case of squared distance, it contains known results. In the case of the Itakura-Saito distance, the distribution of scale parameter characterized by the Itakura-Saito distance is derived, which includes the gamma distribution as a special case. We discuss the possibility of latent bias minimization when the proportion of outliers is large due to the extinction of the bias correction term. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | f-separable distortion measures / Bregman Divergence / Itakura-Saito Distance / Latent Bias / M-estimators |
| Paper # | IT2019-22 |
| Date of Issue | 2019-07-18 (IT) |
| Conference Information | |
| Committee | IT |
|---|---|
| Conference Date | 2019/7/25(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | NATULUCK-Iidabashi-Higashiguchi Ekimaeten |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | freshman session, general |
| Chair | Jun Muramatsu(NTT) |
| Vice Chair | Tadashi Wadayama(Nagoya Inst. of Tech.) |
| Secretary | Tadashi Wadayama(Saga Univ.) |
| Assistant | Hideki Yagi(UEC) |
| Paper Information | |
| Registration To | Technical Committee on Information Theory |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Unbiased Estimation Equation for f-Separable Bregman Distortion Measures and the Properties of Its Estimators |
| Sub Title (in English) | |
| Keyword(1) | f-separable distortion measures |
| Keyword(2) | Bregman Divergence |
| Keyword(3) | Itakura-Saito Distance |
| Keyword(4) | Latent Bias |
| Keyword(5) | M-estimators |
| 1st Author's Name | Masahiro Kobayashi |
| 1st Author's Affiliation | Toyohashi University of Technology(Toyohashi Tech.) |
| 2nd Author's Name | Kazuho Watanabe |
| 2nd Author's Affiliation | Toyohashi University of Technology(Toyohashi Tech.) |
| Date | 2019-07-26 |
| Paper # | IT2019-22 |
| Volume (vol) | vol.119 |
| Number (no) | IT-149 |
| Page | pp.pp.37-42(IT), |
| #Pages | 6 |
| Date of Issue | 2019-07-18 (IT) |