Experimental demonstration of SPM compensation using a complex-valued neural network for 40-Gbit/s optical 16QAM signals

Abstract

We experimentally demonstrate a novel nonlinearity-mitigation scheme based on a complex-valued neural network (CVNN) which is constructed by artificial neurons with complex-valued input and output. The in-phase (I) and quadrature (Q) components of optical signal are operated as complex values in the CVNN. A 40-Gbit/s optical 16QAM signal distorted by SPM was successfully compensated, improving error vector magnitude (EVM) by about 15%. The learning speed of the nonlinear equalizer was improved by using the CVNN, compared with conventional real-valued neural network (RVNN). Furthermore, the study show that CVNN has the potential to improve the computational complexity of RVNN.

Complex-Valued Neural Network Design for Mitigation of Signal Distortions in Optical Links

“…(3), (4). Contrary to the conven-tional real-valued NN architectures [21], in our design we implement complex-valued weights, activation functions, and input symbols as suggested in [22,23] to reflect the complexvalued laws describing the signal propagation. The proposed NN topology is schematically depicted in Fig.…”

Activation functions of artificial-neural-network-based nonlinear equalizers for optical nonlinearity compensation

“…Figure 1(a) shows the construction of a unit in a complex-valued ANN [4]. The complex inner potential, u, is described as…”

“…However, these methods need an enormous amount of calculations, which increases the power consumption and the delay time at the receiver. On the other hand, artificial neural network (ANN)-based nonlinear equalizers are attracting attention because of their lower computational complex-ity [3,4,5,6,7]. In recent years, the rectified linear unit (ReLU) has often been employed as an activation function of the ANN-unit instead of the conventional sigmoid function in deep neural networks (DNNs) used for, e.g., speech recognition and image recognition [8,9].…”

Experimental Verification of Complex-Valued Artificial Neural Network for Nonlinear Equalization in Coherent Optical Communication Systems

“…1. Since the complex-valued symbols, weights and functions are used in our approach, we design our NN as a complex-valued one [15], [16] . We use non-conventional neuron activation functions -|x| 2 , ln(x) and e x to make the NN resembling Eq.…”