Skip to main content

Table 2 Hyperparameters for QSPR/QSAR Model Fine-tuning

From: Inductive transfer learning for molecular activity prediction: Next-Gen QSAR Models with MolPMoFiT

Layer groupsBase Learning RateEpochs
Linear head only3e−24
Linear head + final LSTM layer5e−34
Linear head + final two LSTM layers5e−44
Full model5e−56