Exact component parameter agnostic QoT estimation using spectral data-driven LSTM in optical networks

We propose the use of spectral data-driven LSTM-based machine learning to improve generalized signal-to-noise ratio (gSNR) quality-of-transmission estimation in component parameter-agnostic network scenarios. We show gSNR estimation improvements up to 1.1 dB for unseen networks. © 2021 The Author(s)