Optical nonlinearity compensation using artificial neural-network-based digital signal processing

摘要

We introduce our studies on optical nonlinearity compensation schemes using artificial-neural-networks (ANNs) foroptical communication systems. The ANN-based digital signal processing can compensate nonlinear optical waveformdistortion caused by SPM, XPM, and FWM. We employed real-valued ANNs and complex-valued ANNs, and comparedtheir performances. The performances were studied in comparison with conventional Volterra series transfer function(VSTF). Furthermore, we introduce polarization tracking capability of the ANN for polarization-multiplexed opticaltransmission schemes. We also discuss over-training problem with the ANN and the VSTF.