Table 5 Prediction performance benchmarking for the prediction task of “Transfer Learning on Experimental Data" on the test set.

From: Cross-property deep transfer learning framework for enhanced predictive analytics on small materials data

Property (unit)

Data size

Base

MAE of best SC model

MAE of best TL model

Formation energy (eV/atom)

1643

1.0327

0.0964

0.0708

Band gap (eV)

4920

1.2061

0.4458

0.3551

  1. The table shows the test MAE of the best model selected using Supplementary Table 11 (based on validation MAE) when run on the test set for each of the target materials properties. The SC models are allowed to use PA as input as well, while TL models only use EF-based inputs. The lowest MAE values in each row are highlighted in bold.
Back to article page