Ultralow-complexity fiber nonlinearity compensation based on gradient-driven pruned Kolmogorov-Arnold network.

Fiber nonlinearity compensation is crucial for enhancing the transmission distance and data capacity of wavelength-division multiplexing (WDM) coherent optical communication systems, while existing equalizers based on conventional neural networks fac…

Preview only. Read the full abstract at the source