首页|Optical performance monitoring via domain adversarial adaptation in few-mode fiber
Optical performance monitoring via domain adversarial adaptation in few-mode fiber
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NSTL
Elsevier
We propose and demonstrate an optical performance monitoring (OPM) scheme in few-mode fiber (FMF) based elastic optical networks (EONs). In this work, an adversarial transfer learning (TL) assisted by deep neural networks (DNN), which is used to learn the feature of signal constellations, is investigated for modulation format recognition (MFR) and optical signal-to-noise (OSNR) estimation. Because mode interference affects the projected results in FMF, the DNN is trained to increase link damage tolerance through domain adversarial adaptation (DAA). Five modulation formats have been considered for verifying the feasibility of the proposed scheme, including BPSK, QPSK, 8PSK, 8QAM, and 16QAM. The symbol rate of 12.5 Gbaud is experimentally tested for each scheme across a FMF transmission system. The results reveal that DAA scheme outperforms retraining DNN for requiring one-third training data without any accuracy penalty. Furthermore, the root mean square error (RMSE) of OSNR estimation in this paper can achieve less than 0.1 dB. We anticipate that our results can stimulate further research on the OPM tasks with a feasible deep learning scheme and contribute to the development of FMF-based EONs.
NSTL
Elsevier
