Skip to main content

Table 2 The performance of heteroencoder in both the training and test sets

From: A de novo molecular generation method using latent vector based generative adversarial network

Dataset # compounds Validity (%) Reconstruction error (%)
Training set 974,105 99 18
Test set 10,823 98 20
  1. Percent of valid SMILES strings generated by the decoder (validity), percent of molecules not reconstructed correctly from valid SMILES (reconstruction error)