Table 3.

Energy and density-driven errors of the ML-HK approach on the MD data sets

Molecule	Training trajectories	ΔE		$\Delta {\mathbf{E}}_{\mathbf{D}}^{\mathbf{ML}}$
		MAE	Max	MAE	Max
Benzene	300 K	0.42	1.7	0.32	1.5
	300 + 350 K	0.37	1.8	0.28	1.5
	300 + 400 K	0.47	2.3	0.30	1.8
Ethane	300 K	0.20	1.5	0.17	1.3
	300 + 350 K	0.23	1.4	0.19	1.1
	300 + 400 K	0.14	1.7	0.098	0.62
Malonaldehyde	300 + 350 K	0.27	1.2	0.21	0.74

Errors are given in kcal/mol for different training trajectory combinations