Skip to main content

Table 3 A summary of the final experimental parameters used in both classification experimental designs, in regression Leave-5%-Out internal cross validation, and in designed regression test sets.

From: Classifying and scoring of molecules with the NGN: new datasets, significance tests, and generalization

Parameter Classification Regression 5%-CV Regression Designed
Training Constant (η) 0.60 0.30 0.33
Momentum Coefficient (α) 0.90 0.10 0.66
Convergence Threshold (RMSE) 0.05 0.03 0.04
Initial Random Weight Values [−1.6, −1.0] [−1.6, −1.0] [−1.2, −0.4]
  [1.0, 1.6] [1.0, 1.6] [0.4, 1.2]
Maximum Number of Epochs 5000 7500 7500
SMILES-NGN Hidden Nodes 8 8 8
InChI-NGN Hidden Nodes 8 8 8
Output Scale (0.2, 0.8) [0.2, 0.8] [0.2, 0.8]