TR2021-110

Zero-Multiplier Sparse DNN Equalization for Fiber-Optic QAM Systems with Probabilistic Amplitude Shaping


    •  Koike-Akino, T., Wang, Y., Kojima, K., Parsons, K., Yoshida, T., "Zero-Multiplier Sparse DNN Equalization for Fiber-Optic QAM Systems with Probabilistic Amplitude Shaping", European Conference on Optical Communication (ECOC), September 2021.
      BibTeX TR2021-110 PDF
      • @inproceedings{Koike-Akino2021sep,
      • author = {Koike-Akino, Toshiaki and Wang, Ye and Kojima, Keisuke and Parsons, Kieran and Yoshida, Tsuyoshi},
      • title = {Zero-Multiplier Sparse DNN Equalization for Fiber-Optic QAM Systems with Probabilistic Amplitude Shaping},
      • booktitle = {European Conference on Optical Communication (ECOC)},
      • year = 2021,
      • month = sep,
      • url = {https://www.merl.com/publications/TR2021-110}
      • }
  • MERL Contacts:
  • Research Areas:

    Communications, Machine Learning, Signal Processing

Abstract:

We propose a multiplier-less deep neural network (DNN) to mitigate fiber-nonlinear distortion of shaped constellations. Our DNN achieves an excellent performance-complexity trade-off with progressive lottery ticket hypothesis (LHT) weight pruning and additive powers-of-two (APoT) quantization.