Appendix: Inverse Learning of Symmetries 1 Model 1.1 Implementation

Having generated novel SMILES with potentially identical druglikeliness, we now perform a self-consistency check. That is, we feed the generated SMILES into our model and predict the properties.