Fig. 3: Prediction and inference of noise-driven modulation instability dynamics obtained from numerical simulations.
From: Deep learning prediction of noise-driven nonlinear instabilities in fibre optics
a Example of average spectra obtained after fibre propagation for different dual-seed configurations at the fibre input. The artificial neural network (ANN) prediction (thick blue line) is superimposed on the ground truth given by the simulations (thin black line). b Two-seed wavelength and phase parameters (\({\lambda }_{1}\), \({\lambda }_{2}\) and \({\varDelta \varphi }_{{\mathrm{2,1}}}\)) retrieved by the neural network (white dots) and compared to the seed parameters used in simulations (dashed blue line). The dispersion of the ANN predictions compared to the ground truth (i.e. simulations parameters) is also illustrated via a density plot of the network predictions in each panel background and the root mean square error (RMSE) is provided for each ANN prediction. c Example of average output spectra obtained for different four-seed configurations at the fibre input. Other parameters are the same as in panel (a). d Four-seed wavelength and phase parameters (\({\lambda }_{1}\), \({\lambda }_{2}\), \({\lambda }_{3}\), \({\lambda }_{4}\), \({\varDelta \varphi }_{{\mathrm{2,1}}}\) and \({\varDelta \varphi }_{{\mathrm{4,3}}}\)) retrieved by the neural network (white dots) and compared to the seed parameters used in simulations (dashed blue line). Other parameters are the same as in panel (b).
