Skip to main content
. 2024 Jan 23;64(3):590–596. doi: 10.1021/acs.jcim.3c01777

Table 2. Optimized Hyperparameters for Training of QSAR Neural Networksa.

  QSAR model parameters
Data set Layers Nodes per layer Activation Learning rate
AutoDock Vina 3 256, 256, 32 ReLU, Sigmoid 1 × 10–3
CNS MPO 3 256, 256, 32 ReLU6 1 × 10–3
Aggregation 4 128, 128, 128, 32 ReLU, Sigmoid 5 × 10–4
a

Models were trained for 1000 epochs.