Fig. 3: Model performance and interpretation. | Nature Communications

Fig. 3: Model performance and interpretation.

From: From bulk effective mass to 2D carrier mobility accurate prediction via adversarial transfer learning

Fig. 3

Prediction accuracy on (a) average carrier mobility (defined in Formula 2) under deformation potential theory (DPT); (b) mobility anisotropy (defined in Formula 3); (c) average carrier mobility of transition metal dichalcogenides (TMDs) via solving electron-phonon coupling (EPC) matrix. The model performance is assessed using mean absolute error (MAE) and coefficient of determination (R2), with the corresponding values displayed as text. The feature importance of various feature sets, including adversarial transfer learning learned features (ATL), electronic features (ELE), symmetry features (SYM), and geometry features (GEO), is analyzed for models predicting (d) average DPT carrier mobility, (e) mobility anisotropy, and (f) average EPC carrier mobility of TMDs. The orange and blue colors represent electrons and holes, respectively, and the corresponding features in each feature set can be found in Supplementary Table 2. Source data are provided as Source Data files.

Back to article page