Figure 4
From: Deep learning based analysis of microstructured materials for thermal radiation control
MAE for the transmission and reflection predictions compared to FDTD simulations for the 23 unseen library materials. (a) Plotted error when the materials are completely “unseen” and (b) after 5 simulations for each material are included in the training/testing/validation process. The log and then linearly normalized average extinction coefficient is shown in the z-axis, pointing to the role of the material in predicting where the error will occur. The error’s (x,y) distance from an MAE error of zero is shown with the color bar. Including 5 simulations systematically reduces the prediction error for the rest of the dataset, indicating that very little data is needed to calibrate the model for new materials and lead to accurate predictions.
