Convolutional Neural Networks with Multiple Layers per Span for Nonlinearity Mitigation in Long-Haul WDM Transmission Systems

Overcoming fiber nonlinearity is one of the most challenging tasks in optical fiber links and it is a major limiting factor for extending their capacity. Digital backward propagation (DBP) method can be used to mitigate nonlinear transmission impairments [1] , but its complexity prevents any real-time implementation in these systems. On the other hand, it has been recently shown that deep neural networks can provide a good approximation of DBP at lower computational cost [2] . In this work, we continue the investigation of the proposed deep convolutional neural network (DCNN) [3] for long-haul WDM transmission systems. We study the effect of the number of neural network layers on the efficiency of nonlinear distortion compensation.