光纤传输实验中用于非线性抑制的神经网络少比特量化
信号处理
2022-05-26 v2
摘要
在一个16-QAM 9x50km双偏振光纤传输实验中，对神经网络进行量化以抑制非线性和器件失真。训练后加性2的幂次6比特量化带来的Q因子损耗可忽略。在5比特时，模型尺寸缩减85%，且仅有0.8 dB损耗。
引用
@article{arxiv.2205.11284,
title = {Few-bit Quantization of Neural Networks for Nonlinearity Mitigation in a Fiber Transmission Experiment},
author = {Jamal Darweesh and Nelson Costa and Antonio Napoli and Bernhard Spinnler and Yves Jaouen and Mansoor Yousefi and .},
journal= {arXiv preprint arXiv:2205.11284},
year = {2022}
}
备注
4 pages ,3 figuers