Abstract As optical communication systems have developed quickly, more sophisticated methods that were previously only used for long-distance transmissions having progressively navigated into systems with faster speeds are needed for a variety of applications, including mobile front haul and optical access networks. Machine learning (ML) is a technique that has garnered significant attention in the field of short-optical reach communications for applications such as passive optical network (PON), indoor wireless communications, and signal processing due to its adaptability, high accuracy, and implementation efficiency. This work focuses on ML and neural networks (NN) techniques that is applicable to complicated systems and optimization. The datasets for different classifiers and classification features need to be trained and tested utilizing raw ensemble predictions, a calibration phase starts after the raw ensemble predictions are finished in order to assess the accuracy of the predictions made in the preceding phase and make any necessary improvements. The proposed ML methods use a database, which was validated utilizing experimental derived data from the TWDM-PON system to estimate the Q-factor, logarithm of bit error rate (Log(BER)), and receiver sensitivity. The classifier accuracy of the mean absolute error (MAE), mean squared error (MSE), and the coefficient of determination (R 2 ) results from simulations are performed utilizing Python.
Ridha et al. (Wed,) studied this question.