Fig. 2: Relative errors for Training and Test Set 1. | Communications Chemistry

Fig. 2: Relative errors for Training and Test Set 1.

From: Data-driven analysis of the number of Lennard–Jones types needed in a force field

Fig. 2

Relative errors for experimental data (heats of vaporization, densities, dielectric constants), and for the ForceBalance objective function, for Training Set 1 (red) and Test Set 1 (blue). The test set objective functions are scaled by 1/15 (see main text). The number of parameters may be deduced from names of the LJ models; e.g., HC3ON has 1 + 3 + 1 + 1 = 6 parameters.

Back to article page