Spectral and Spatial Power Evolution Design With Machine Learning-Enabled Raman Amplification

We present a machine learning (ML) framework for designing desired signal power profiles over the spectral and spatial domains in fiber span. The proposed adjusts Raman pump values to obtain two-dimensional (2D) using convolutional neural network (CNN) followed by differential evolution (DE) technique. CNN learns mapping between 2D their corresponding data-set generated exciting amplification setup. Nonetheless, its performance is not accurate of practical interest, such as flat or symmetric (with respect middle point distance). To adjust more accurately, DE fine-tunes initialized design profile with lower cost value. In fine-tuning process, employs direct model which consists 8 bidirectional propagating pumps, including 2 s-order 6 first order, an 80 km evaluate broadband profiles, two goals wavelength division multiplexing (WDM) system performing whole C-band. Results indicate framework’s ability achieve maximum excursion 2.81 dB flat, asymmetry 14% profile.