Skip to main content

Table 1 Benchmark results of the general model G

From: Conditional reduction of the loss value versus reinforcement learning for biassing a de-novo drug design generator

Temperature

Validity

Novelty

Uniqueness

intDiv

1

34.2 ± 5.1

96.5 ± 0.9

99.8 ± 0.7

0.84 ± 0.05

0.80

52.4 ± 4.5

94.2 ± 1.2

98.0 ± 1.8

0.83 ± 0.18

0.60

65.8 ± 3.6

93.3 ± 1.9

97.4 ± 3.2

0.82 ± 0.90

0.50

80.0 ± 2.9

92.5 ± 2.5

88.2 ± 3.9

0.81 ± 1.92

0.40

85.7 ± 1.8

91.8 ± 3.7

65.7 ± 5.0

0.79 ± 2.02

0.20

99.7 ± 0.2

91.0 ± 5.2

11.1 ± 6.8

0.73 ± 3.26