| Presentation | 2023-03-14 An extension of the Sinkhorn divergence via convex combination of the bias correction term Kyohei Miwa, Yu Nishiyama, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Sinkhorn divergence (SD) is a divergence in which a bias correction term is added to the entropic optimal transport (EOT) so that EOT cost between $2$ probability distributions $alpha,beta$ satisfies the conditions of divergence (non-negativity and non-degeneracy). However, the bias correction term proposed in the previous studies is not a necessary condition for satisfying the conditions of divergence. In this study, we propose an extended SD that generalizes the bias correction term of SD to the convex combination. The proposed method satisfies the condition of divergence and inherits the properties of SD such as convexity and symmetry. We show a numerical example where a probability distribution is transported via a more efficient path than the conventional method (SD minimization) in a simple optimization setting among $1$-dimensional distributions. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | optimal transport / entropic optimal transport / Sinkhorn divergence / extended Sinkhorn divergence / convex combination |
| Paper # | NC2022-100 |
| Date of Issue | 2023-03-06 (NC) |
| Conference Information | |
| Committee | NC / MBE |
|---|---|
| Conference Date | 2023/3/13(3days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | The Univ. of Electro-Communications |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | Brain architecture, General |
| Chair | Hiroshi Yamakawa(Univ of Tokyo) / Junichi Hori(Niigata Univ.) |
| Vice Chair | Hirokazu Tanaka(Tokyo City Univ.) / Hisashi Yoshida(Kinki Univ.) |
| Secretary | Hirokazu Tanaka(NTT) / Hisashi Yoshida(NICT) |
| Assistant | Yoshimasa Tawatsuji(Waseda Univ.) / Tomoki Kurikawa(KMU) / Emi Yuda(Tohoku Univ) / Miki Kaneko(Osaka Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Neurocomputing / Technical Committee on ME and Bio Cybernetics |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | An extension of the Sinkhorn divergence via convex combination of the bias correction term |
| Sub Title (in English) | |
| Keyword(1) | optimal transport |
| Keyword(2) | entropic optimal transport |
| Keyword(3) | Sinkhorn divergence |
| Keyword(4) | extended Sinkhorn divergence |
| Keyword(5) | convex combination |
| 1st Author's Name | Kyohei Miwa |
| 1st Author's Affiliation | The University of Electro-Communications(UEC) |
| 2nd Author's Name | Yu Nishiyama |
| 2nd Author's Affiliation | The University of Electro-Communications(UEC) |
| Date | 2023-03-14 |
| Paper # | NC2022-100 |
| Volume (vol) | vol.122 |
| Number (no) | NC-425 |
| Page | pp.pp.48-53(NC), |
| #Pages | 6 |
| Date of Issue | 2023-03-06 (NC) |