Egor Sedov

Egor Sedov

PhD | Researcher | Machine Learning | AI engineering

Neural Networks For Nonlinear Fourier Spectrum Computation

Published in 2021 European Conference on Optical Communication (ECOC) | IEEE, 2021

We demonstrate that neural networks can outperform conventional numerical nonlinear Fourier transform algorithms for processing the noise-corrupted optical signal. Applying the Bayesian hyper-parameters optimisation, we design the architecture of neural networks capable to compute nonlinear signal spectrum at low SNR more accurately than conventional algorithms.

Download paper here or here

@inproceedings{sedov2021neural,
  title={Neural Networks For Nonlinear Fourier Spectrum Computation},
  author={Sedov, Egor and Freire, Pedro J and Chekhovskoy, Igor and Turitsyn, Sergei and Prilepsky, Jaroslaw},
  booktitle={2021 European Conference on Optical Communication (ECOC)},
  pages={1--4},
  year={2021},
  organization={IEEE}
}