Complex Fully Connected Neural Networks for Nonlinearity Compensation in Long-Haul Transmission Systems

The complex-valued fully connected neural networks are applied for nonlinearity compensation in fiber optic communication systems. The superiority of a such approach over the real-valued neural networks and linear compensation schemes is demonstrated.