Skip to main content

Table 6 Among all the starting molecules in Test-Original, the percentage of starting molecules for which each model generates either less or more desirable molecules when compared with the other two models

From: Molecular optimization by capturing chemist’s intuition using deep neural networks

  Less than the others (%) More than the others (%)
HierG2G 46.50 10.02
Seq2Seq 23.16 21.81
Transformer 8.89 46.20