| Presentation | 2022-08-29 Compensation Performance of DNN-based Nonlinear Equalizer for Optical Communication Systems Jinya Nakamura, Kai Ikuta, Daisuke Motai, Moriya Nakamura, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | We investigated and compared the performances of three-layer-ANN- and four-layer-DNN-based nonlinear equalizers used for the compensation of optical nonlinear distortion. We confirmed that the learning speed of DNN was faster than that of ANN when we optimized the learning rate. However, the learning process of the DNN-based nonlinear equalizer was unstable and diverged in some cases. We clarified that the divergence of the learning was caused by the local-minimum trapping where the weights became extremely large and noise is clipped by the saturation characteristics of the DNN units. Furthermore, we confirmed that the divergence can be suppressed by using L2 regularization. When the compensation accuracies of the ANN and DNN were compared based on computational complexity, no significant difference was observed. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | optical nonlinear compensation / nonlinear equalizer / deep neural network / regularization |
| Paper # | PN2022-10 |
| Date of Issue | 2022-08-22 (PN) |
| Conference Information | |
| Committee | PN |
|---|---|
| Conference Date | 2022/8/29(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Hideaki Furukawa(NICT) |
| Vice Chair | Kohei Shiomoto(Tokyo City Univ.) / Ryuichi Sugizaki(Furukawa Electric) / Noboru Yoshikane(KDDI Research) |
| Secretary | Kohei Shiomoto(NTT) / Ryuichi Sugizaki(Univ. of Electr-Comm.) / Noboru Yoshikane |
| Assistant | Kenji Ishii(Mitsubishi Electric) / Yojiro Mori(Nagoya Univ.) |
| Paper Information | |
| Registration To | Technical Committee on Photonic Network |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Compensation Performance of DNN-based Nonlinear Equalizer for Optical Communication Systems |
| Sub Title (in English) | |
| Keyword(1) | optical nonlinear compensation |
| Keyword(2) | nonlinear equalizer |
| Keyword(3) | deep neural network |
| Keyword(4) | regularization |
| 1st Author's Name | Jinya Nakamura |
| 1st Author's Affiliation | Meiji University(Meiji Univ.) |
| 2nd Author's Name | Kai Ikuta |
| 2nd Author's Affiliation | Meiji University(Meiji Univ.) |
| 3rd Author's Name | Daisuke Motai |
| 3rd Author's Affiliation | Meiji University(Meiji Univ.) |
| 4th Author's Name | Moriya Nakamura |
| 4th Author's Affiliation | Meiji University(Meiji Univ.) |
| Date | 2022-08-29 |
| Paper # | PN2022-10 |
| Volume (vol) | vol.122 |
| Number (no) | PN-169 |
| Page | pp.pp.10-14(PN), |
| #Pages | 5 |
| Date of Issue | 2022-08-22 (PN) |