Table 1 Comparison of the MAEs between several benchmarked models and HVNet trained on MD17 dataset using 1000 training samples (energies in meV and forces in meV Å−1).
From: Heterogeneous relational message passing networks for molecular dynamics simulations
SchNeta | DimeNetb | PAINNc | HVNet | ||
|---|---|---|---|---|---|
Benzene | Energy | 3.44 | 3.354 | – | 0.319 |
Force | 13.33 | 8.041 | – | 1.223 | |
Toluene | Energy | 5.16 | 4.386 | 4.171 | 1.967 |
Force | 24.51 | 9.288 | 4.386 | 5.268 | |
Malonaldehyde | Energy | 5.59 | 4.472 | 3.913 | 1.422 |
Force | 28.38 | 16.469 | 13.717 | 6.215 | |
Salicylic acid | Energy | 8.60 | 5.762 | 4.902 | 4.128 |
Force | 36.55 | 16.082 | 8.987 | 10.887 | |
Aspirin | Energy | 15.91 | 8.772 | 6.837 | 9.935 |
Force | 58.05 | 21.457 | 15.953 | 11.734 | |
Ethanol | Energy | 3.44 | 2.752 | 2.709 | 1.258 |
Force | 16.77 | 9.89 | 9.89 | 6.102 | |
Uracil | Energy | 6.02 | 4.945 | 4.472 | 1.660 |
Force | 24.08 | 12.943 | 6.02 | 3.999 | |
Naphtalene | Energy | 6.88 | 5.246 | 5.031 | 2.728 |
Force | 24.94 | 9.245 | 3.569 | 4.469 |
