Table 9 The hyperparameter ranges in the GCN, where the values selected for the best model are shown in bold

From: Molecular optimization using a conditional transformer for reaction-aware compound exploration with reinforcement learning

Hyperparameters

Search range and best value

Dimension

128, 256, 512

Convolution Layer

1, 2, 3, 4, 5

Dense Layer

1, 2, 3, 4, 5

Learning rate

0.00001 – 0.1 (0.0004)

Back to article page