Reduction of Nonlinear Distortion in SOA-Based Wavelength Conversion System by Post-Blind-Compensation based on Machine Learning Clustering

Nonlinearities in SOA-based wavelength conversion are experimentally tackled using post-blind-compensation based on machine learning clustering for coherent 16/64-QAM signals. We show that K-means outperforms fuzzy-logic since it tackles more effectively non-circularly-symmetric Gaussian noise and nonlinearity.