Figure 7
From: Real-time reconstruction of high energy, ultrafast laser pulses using deep learning
Validation loss of SHERPA trials, measured by the mean squared error, over time. Each line depicts the validation loss of a different SHERPA trial during the course of training of the simulated few-cycle networks. (a) Trials from phase networks with varying hyperparameters. (b) Trials from fluence networks with varying hyperparameters. Note: not all 500 trials are shown in each figure. Some trials with higher validation losses are left out for figure clarity. This discards 50 and 145 networks for a and b, respectively.
