Towards intelligent fiber laser design by using a feed-forward neural network

We demonstrated a high accuracy prediction of the fiber laser output parameters by using a feed-forward neural network. We explored both the gain and spectral filter parameters to test the prediction performance of the neural network and realized the mapping between cavity parameters and laser output performance. We also investigated how the number of hidden layers could influence the accuracy of prediction. Based on the results, the output spectrum and temporal pulse profiles can be predicted with high accuracy in various fiber laser designs. Our work paves the way to intelligent laser design with ultimate autonomy.