Abstract

We use a neural network trained jointly by multi-task learning on datasets acquired at multiple wavelengths to mitigate the impact of chromatic dispersion in 4×200Gb/s CWDM4 PAM4 transmission. By sharing a single set of weights among all involved wavelengths, while keeping the biases reconfigurable, we enable logic simplification of multipliers in the VLSI implementation of the neural network. Results show that the neural network equalizer achieves a similar BER compared with a Volterra equalizer with 71% reduction in hardware area.

© 2023 The Author(s)