fig3

Figure 3. Comparison of the predicted atomistic force and the DFT counterpart of monolayer (A) GeSe and (B) PbTe. The testing dataset contains ~0.86 million and ~1.5 million points for GeSe and PbTe, respectively. We calculate the RMSE, MAE, and R² of the ML model for the testing data, and only 10% of the data are shown in the plots. DFT: Density functional theory; MAE: mean absolute error; ML: machine learning; RMSE: root mean square errors.