Table 2 Generation performance of ConditionCDVAE+ and baseline models on MP-20 and J2DH-8 datasets. Best results are marked in bold.
From: Deep generative model for the inverse design of Van der Waals heterostructures
Dataset | Method | Validity (%) \(\uparrow\) | COV (%) \(\uparrow\) | Property \(\downarrow\) | |||
|---|---|---|---|---|---|---|---|
Struc. | Comp. | R. | P. | \(\rho\) | #elem. | ||
MP-20 | FTCP21 | 1.55 | 48.37 | 4.72 | 0.09 | 23.71 | 0.7363 |
DiffCSP27 | 100.0 | 83.25 | 99.71 | 99.76 | 0.3502 | 0.3398 | |
CDVAE21 | 100.0 | 86.70 | 99.15 | 99.49 | 0.6875 | 1.432 | |
DP-CDVAE28 | 99.59 | 85.44 | 98.93 | 98.96 | 0.4037 | 0.9179 | |
ConditionCDVAE+ | 99.91 | 82.43 | 99.59 | 99.37 | 0.3426 | 0.2118 | |
J2DH-8 | FTCP | 88.47 | 87.42 | 13.32 | 0.00 | 3.165 | 0.4175 |
DiffCSP | 37.81 | 99.94 | 100.00 | 13.51 | 0.9699 | 0.4355 | |
CDVAE | 100.00 | 100.00 | 100.00 | 80.84 | 0.1040 | 0.2404 | |
DP-CDVAE | 100.00 | 99.50 | 100.00 | 68.30 | 0.1606 | 0.0604 | |
ConditionCDVAE+ | 100.00 | 100.00 | 100.00 | 87.41 | 0.1067 | 0.0964 | |
