Gated Recurrent Neural Networks based Pre- Distortion for Digital-to-Analog Converter

Coherent optical transmission at high baud-rates suffers from the steep roll-off of digital-to-analog converters (DACs) at high frequency. We propose gated recurrent neural networks as digital pre-distortors to compensate DAC response at 120Gbaud. We demonstrate experimentally an improvement of 1.2 dB over linear pre-distortion.