Performance Analysis of DP-64-QAM Optical Networks Using Convolutional Neural Network

This study discusses a convolutional neural network established perturbative nonlinearity compensation that represents two channels with first-order perturbation theory that builds a classifier as well as a regressor. We need to represent the convolutional neural network equalizer in 370 kilometer 120-Gb/s dual-polarization 64-quadrature-amplitude modulation fiber-optic networks. Also, well defined the nonlinear activation function for convolutional neural network equalizer. This research work calculates the bit error rate and launched optical signals. If the channel size is 10, the bit error rate performance for the convolutional neural network classifier is 0.0011 (1 dBm) and the convolutional neural network regressor is 0.0018 (0 dBm). The bit-error ratio is fewer than 6%, and a threshold value is 0.0037 (−3 dBm to 3 dBm). If the channel size is 14, the bit-error-ratio (−4 dBm, 4 dBm, and 5 dBm) is fewer than 0.0018. We are required to analyze different network complexity by using convolutional neural network classifiers with regressors. We need to simulate the time complexity up to 24.99%, and the space complexity is 49.99%.