Table 1.

Performance comparison for unconditional molecular graph generation

Dataset	Metric	GraphVAE [28]	MolGAN [29]	Proposed
				A1	A2	Full
QM9	Validity	0.610	0.981	0.550	0.965	0.945
	Uniqueness	0.409	0.104	0.293	0.275	0.343
	Novelty	0.850	0.942	0.612	0.737	0.806
	G-mean	0.596	0.458	0.462	0.581	0.639
ZINC	Validity	0.140	0.017	0.008	0.926	0.919
	Uniqueness	0.316	0.201	0.949	0.614	0.762
	Novelty	1.000	1.000	1.000	1.000	1.000
	G-mean	0.354	0.151	0.197	0.828	0.888

Best score for each metric is given in italic