Fig. 14: Models with Transfer Learning (TL) and without Transfer Learning (Non-TL).

a, c Root mean square error (RMSE) on the test dataset and b, d Computational time to generate the training data. In the case of aluminum (a, b), the TL model is trained using 32 and 108 atom data. For SiGe (c, d), the TL model was trained using 64 and 216 atom data. In the case of aluminum, the non-TL model is trained using 108 atom data. Whereas, in the case of SiGe, the non-TL model is trained using 216 atom data.