Fig. 1. An example of the wavefront validation with three different reconstruction methods: LSF, SVD, and ANN^[9]

Fig. 2. Comparison of outputs of ISNet, optimized least squares (OLS), and Southwell algorithms at three levels of turbulence^[11]

Fig. 3. Illustration of convolutional neural network architectures^[12]

Fig. 4. Architecture of the neural network and process of model training and testing^[13]

Fig. 5. Modified U-net architecture^[17]

Fig. 6. Percentage of cases with residual RMS WFE below 1/10 of the Marechal criterion when using random starting points and the CNN’s predictions^[18]

Fig. 7. CNN architectures^[24]

Fig. 8. Sketch map of the feature-based phase retrieval wavefront sensing approach using machine learning^[25]

Fig. 9. Data flow of the training and estimation processes^[27]

Fig. 10. Comparison of the four models’ simulation results under strong turbulence and weak turbulence^[37]

Fig. 11. Architecture for the wavefront prediction network^[10]