Accurate Evaluation of Cascaded Four-wave Mixing Products Generation in Fiber with Optical Feedback Based on Multilayer Perceptron

Abstract

The introduction of optical feedback mechanism enhances the cascaded four-wave mixing (CFWM) effect in high nonlinear fiber, and the machine learning algorithm is used to optimize and predict the bandwidth and number of the CFWM products through controlling parameters in high nonlinear fiber, such as fiber length, pump power and feedback coefficient. Compared with the single pass situation, the results show that the bandwidth is increased to 300 nm with up to 41 products, and the products number is also improved by introducing optical feedback strategy. The comparison between the simulation results and machine learning results demonstrated that the neural network model has the potential for analyzing and predicting the CFWM products in the high nonlinear fiber with a feedback system. The model is characterized by MSE below 0.01 and improves the time efficiency by 81.9%. This research can pave the way for realizing cross-over studies bridge between nonlinear fiber optics and machining learning.