Skip to main content

Table 3 Predictive performance of the SVM models

From: Memory-assisted reinforcement learning for diverse molecular de novo design

Target Set Bal. ACC ROC AUC F1 MCC
HTR1A Training 0.98 0.99 0.77 0.78
Validation 0.96 0.99 0.75 0.75
Test 0.96 0.99 0.75 0.75
DRD2 Training 0.99 0.99 0.77 0.79
Validation 0.93 0.98 0.70 0.71
Test 0.95 0.99 0.71 0.72
  1. The SVM model was trained on the training set and the hyperparameters c, the choice of the kernel, and the class weight were optimized towards a high F1 score on the validation set. The test set was used to estimate the predictive performance of unknown compounds. “bal” stands for balanced