Nonlinearity mitigation with neural networks in vector mm-wave system.

Abstract In this paper, a Volterra series nonlinear equalization (NLE) scheme and two kinds of neural network equalization (NNE) schemes are presented to compensate the nonlinear distortions in vector mm-wave system over fiber link. The principle of NLE, two-layer forward neural network (FNN) and radial basis function neural network (RNN) scheme are presented and their performance are compared through a 5Gbaud 16QAM over 40 GHz RF carrier simulation system. The presented NLE and NNE scheme are both proved to be feasible in nonlinearity mitigation in vector mm-wave system, and NNE scheme outperforms the conventional nonlinear equalization using statistics algorithm. Highlights • Volterra based NLE, FNN and RNN are investigated in vector mm-wave system. • FNN and RNN shows similar performance and feasibility for nonlinearity compensation. • NN class schemes outperform the conventional NLE using statistics algorithm. [ABSTRACT FROM AUTHOR]