| Presentation | 2021-01-14 Automatic design of constellations and discriminant boundary by auto-encoder type machine learning Yudai Goto, Masanori Hanawa, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Autoencoder (AE), a type of machine learning, was applied to the automatic design of signal constellations and discriminant boundary for coherent optical fiber transmission systems. A realistic single-wavelength, point-to-point single-mode fiber coherent optical transmission system with considering both wavelength dispersion and self-phase modulation is inserted in the middle of the AE. It is shown that the transmitter of the AE automatically constructs an encoder with a constellation map corresponding to noise, wavelength dispersion and self-phase modulation given during AE learning, and the receiver automatically constructs a decoder with a linear discriminant boundary. It is also shown that setting an appropriate signal-to-noise power ratio (SNR) during learning stabilizes the learning results and improves the BER characteristics, and that the BER characteristics obtained by AE are close to those of 16QAM with wavelength dispersion and self-phase modulation compensated by the back propagation method. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | Coherent transmission / Machine learning / Autoencoder / Design of constellations / Discriminant boundary |
| Paper # | OCS2020-32 |
| Date of Issue | 2021-01-07 (OCS) |
| Conference Information | |
| Committee | OCS / CS |
|---|---|
| Conference Date | 2021/1/14(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | Joji Maeda(Tokyo Univ. of Science) / Jun Terada(NTT) |
| Vice Chair | / Daisuke Umehara(Kyoto Inst. of Tech.) |
| Secretary | (NTT) / Daisuke Umehara(Fujikura) |
| Assistant | / Hiroyuki Saito(OKI) / Takahiro Yamaura(Toshiba) |
| Paper Information | |
| Registration To | Technical Committee on Optical Communication Systems / Technical Committee on Communication Systems |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Automatic design of constellations and discriminant boundary by auto-encoder type machine learning |
| Sub Title (in English) | |
| Keyword(1) | Coherent transmission |
| Keyword(2) | Machine learning |
| Keyword(3) | Autoencoder |
| Keyword(4) | Design of constellations |
| Keyword(5) | Discriminant boundary |
| 1st Author's Name | Yudai Goto |
| 1st Author's Affiliation | University of Yamanashi(Univ. of Yamanashi) |
| 2nd Author's Name | Masanori Hanawa |
| 2nd Author's Affiliation | University of Yamanashi(Univ. of Yamanashi) |
| Date | 2021-01-14 |
| Paper # | OCS2020-32 |
| Volume (vol) | vol.120 |
| Number (no) | OCS-309 |
| Page | pp.pp.27-32(OCS), |
| #Pages | 6 |
| Date of Issue | 2021-01-07 (OCS) |