Nonlinear Neural Network Equalizer for Metro Optical Fiber Communication Systems

We present a neural network-based nonlinear electronic feed-forward equalizer. It compensates for the chromatic dispersion (CD) distortions in fiber optic communication systems with direct photo-detection. The proposed equalizer achieves bit error rate (BER) performance comparable to the maximum-likelihood sequence estimator (MLSE), with significantly lower computational cost. The complexity of the introduced equalizer scales linearly with the length of the intersymbol interference (ISI) as opposed to exponential growth the MLSE complexity.