We report a neural-network based erbium-doped fiber amplifier (EDFA) gain model built from experimental measurements. The model shows low gain-prediction error for both the same device used for training (MSE leq 0.04 dB^2) and different physical units of the same make (generalization MSEleq 0.06 dB^2).
|Title of host publication||Proceedings of 2020 European Conference on Optical Communications|
|Number of pages||4|
|Publication date||Dec 2020|
|Publication status||Published - Dec 2020|
|Event||46th European Conference on Optical Communication - Virtual event, Brussels, Belgium|
Duration: 6 Dec 2020 → 10 Dec 2020
Conference number: 46
|Conference||46th European Conference on Optical Communication|
|Period||06/12/2020 → 10/12/2020|
Bibliographical noteFunding Information:
This work is supported by the Villum Foundations (VYI grant OPTIC-AI no.29344), the EU H2020 programme (Marie Skłodowska-Curie grant no. 754462) and the DNRF CoE SPOC (ref. DNRF123).
© 2020 IEEE.