We discuss our recent work on machine learning based nonlinear equalization in long haul transmission sytems. We show that dynamic multi-perceptron networks can deal with the memory properties of the fibre channel and provide efficient mitigation of nonlinear impairments at lower computational cost when compared to conventional digital back propagation methods.
|Title of host publication||Signal Processing in Photonic Communications 2019|
|Publisher||Optical Society of America|
|Publication status||Published - 29 Jul 2019|