Fig. 6: Glass properties predicted by the GBM-LASSO model validated with experimental results.
From: Predicting densities and elastic moduli of SiO2-based glasses by machine learning
Experimental data were collected from the Sciglass 7.12 database56. The GBM-LASSO model is only trained with the MD training set and the experimental data were not used for training. a Shear modulus; b Young’s modulus; c Density. The dashed line is the identity where the predictions are equal to the experimental values. The hexagonal unit with a hotter color means that there are more data points within the coverage area of the unit. The dashed-line circles in a mark out typical examples of prediction outliers caused by experimental data inconsistency.
