Estimação Precisa da Qualidade de Transmissão com Redes Neurais Profundas para Sistemas Ópticos Densos

The advent of new Internet technologies and ap- plications has increased the need for high-speed and reliable data transport. DWDM optical networks use Transmission Qua- lity (QoT) metrics, such as Generalized Signal-to-Noise Ratio (GSNR), for channel provisioning. This parameter requires computationally expensive analytical prediction. Thus, this article proposes a model of deep neural networks to estimate the Gene- ralized Signal-to-Noise Ratio accurately and quickly. The model is validated through simulations, which in turn are validated with data from a real network in operation. The results show an absolute percentage error of less than 0.54% between the estimation obtained by the proposed model and the simulated value.