Figure 2

Impact of input representation on the accuracy of BRNet. The x-axis shows the dataset size on a log scale, and the y-axis shows the percentage change in MAE of the model trained using composition-based elemental fraction as input w.r.t. the model trained using composition-based physical attributes as input (calculated as ((MAE\(_{EF}\)/MAE\(_{PA})-1\))Â \(\times\)Â 100\(\%\)).