Artificial intelligence can predict complex ultrafast dynamics in optical fibers

We show how a ‘recurrent neural network’ that possesses an internal memory can predict nonlinear dynamics over extended distances in optical fibers for a wide range of scenarios, from higher-order soliton compression to octave-spanning supercontinuum generation and even multimode nonlinear propagation. Once trained the network can predict spatio-temporal or spatio-spectal evolution maps in milliseconds. Check our paper in Nature Machine Intelligence